MODERN COLOUR THEORY FOR

TRADITIONAL AND DIGITAL PAINTING

MEDIA
by Dr David. J.C. Briggs,
Julian Ashton Art School and National Art School, Sydney, Australia.
Chairperson, NSW Division, Colour Society of Australia.

 Site contents
 Acknowledgements

Colours of twenty two common artists pigments at various thicknesses

over a white ground. Photographed colours displayed in YCbCr space
using the program ColorSpace by Philippe Colantoni.

This website presents an account of the dimensions of colour and light

perception, written for painters using either traditional or digital media. The
conceptual framework presented here was developed as a component of Colour,
Light and Vision, a course in colour theory and practice for artists that I have been
presenting since 1998 at the Julian Ashton Art School, Sydney, of Theories of Colour, a
lecture course on the history of colour theory and practice that I presented at
the National Art School, Sydney, in 2009-2011, and of ongoing practical painting classes at both
schools.

Site Contents
All pages published in 2007 unless otherwise stated; latest revisions indicated in red:

Home
1 The Dimensions Introduced (revised Jan-Feb 2017)

 1.1 Colours in Space

 1.2 The Dimensions of What, Exactly? (added 15/1/17)
 1.3 The Dimensions of Colour: Lightness (added 16/1/17)
 1.4 The Dimensions of Colour: Hue (added 21/1/17)
 1.5 The Dimensions of Colour: Chroma (added 23/1/17)
 1.6 The Dimensions of Colour: Brightness and
Colourfulness (added 30/1/17)
 1.7 The Dimensions of Colour: Saturation (added 31/1/17)
 1.8 The Dimensions of Colour: Blackness and
Brilliance (added 15/2/17)

2 Basics of Light and Shade

 2.1 Specular and Diffuse Reflection

 2.2 The Zone of Light
 2.3 The Zone of Shadow

3 Basics of Colour Vision

 3.1 Introduction
 3.2 Trichomacy and Opponency (revised 08/10/11)
 3.2 Adaptation and Successive Contrast
 3.4 Colour Constancy
 3.5 Simultaneous Contrast and Assimilation
 3.6 What is Colour? (added 09/3/14)
 3.7 Answers to "What is Colour"? (added 09/3/14)

4 Additive Colour Mixing

 4.1 Additive Primaries (revised 5/8/12)

 4.2 Additive Mixtures (revised 5/8/12)
 4.3 Additive Complementaries (revised 5/8/12)
 4.4 Additive-Averaging Mixing (revised 5/8/12)
 4.5 Object Colours (revised 5/8/12)

5 Subtractive Colour Mixing

 5.1 Subtractive Mixing (revised 13/8/12)

 5.2 Ideal Subtractive Primaries (revised 12/8/12)
 5.3 Subtractive Complementaries (revised 26/6/12)

6 Colour Mixing in Paints

 6.1 Mixing Paints (revised 2/11/12)

  6.2 Primary Colours (revised 2/11/12)
 6.3 Paint-mixing Principles (revised 2/11/12)

7 Hue

 7.1 Hue from Aristotle to Newton (revised 15/4/13)

 7.2 The RYB Hue Circle or Artist's Colour Wheel (revised
15/4/13)
 7.3 Hue Systems Based on Opponent Colours (revised
28/7/13)
 7.4 Hue Systems Based on Additive Complementaries
 7.5 Hue Systems Based on Pigment-Mixing
Complementaries
 7.6 Orthogonal Systems
 7.7 Warm and Cool Hues

8 Lightness and Chroma

 8.1 Lightness
 8.2 Chroma
 8.3 Hue-Chroma-Lightness Colour Spaces

9 Brightness and Saturation

 9.1 Brightness, Saturation and Colorfulness

 9.2 RGB, CMY and CMYK Colour Space
 9.3 HSB (=HSV) Colour Space
 9.4 HLS (=HSL) and HSI Colour Space

10 Principles of Colour

 10.1 Shading Series (revised 26/7/14)

 10.2 Consistency of Relative Brightness
 10.3 The Scale of Brilliance
 10.4 Effect of Coloured Illumination
 10.5 Effect of Multiple Light Sources
 10.6 Effect of Distance From Light
 10.7 Effect of Inclination to Light
 10.8 Effects of Atmosphere
 10.9 Applying the Principles in Paint

11 References

12 Glossary (under construction)

Epilogue: Traditional and Modern Colour Theory (added
30/6/15
(Now also on
YouTube!: https://www.youtube.com/watch?v=4enFjTGVTnc )

 Modern Colour Theory (added 30/6/15)

 Traditional Colour Theory Strikes Back! (added. 30/6/15)

Acknowledgements
Warm thanks go to the "Dimensions of Colour team", Xavier Peria, Ray Kristanto, Noopur Patel,
Atania Trinata, and Debolina Bandyopadhyay from the 2007 second year Multimedia course at
the Billy Blue School of Graphic Arts, Sydney, and their teacher Dave Agius, for creating the site,
including the interactive animations. Thanks also to Ben Green for generously hosting the site
during its first year online, and to ibiblio, "the public's library and digital archive" at the
University of North Carolina - Chapel Hill, for accepting the site into their collection and hosting
it since then. Finally, thanks to all of those who have added links to this site on their websites,
blogs and forum posts, and especially to the following for their published comments:

Mark Fairchild (USA), Professor of Color Science & Imaging Science, Rochester Institute of
Technology, New York, and author of the textbook Color Appearance
Models (Wiley): Essentially an online textbook/tutorial on appearance, or "the dimensions of
colour and light" written from the perspective of artists. The site is very nicely done and blends
technical and artistic information well.

James Gurney (USA), illustrator, fine artist, author of numerous books including
the Dinotopia series and Color and Light: A Guide for the Realist Painter: David Briggs is none
other than the mastermind behind the website "Dimensions of Color." It's one of the best
resources on light and color on the Internet. I owe much of what I've learned on the topic to
[Dr] Briggs. ...His website "HueValueChroma" has that rare combination of depth and clarity.

David Gray (USA), fine artist and painting teacher: ... an absolutely indispensable source of
color knowledge for the realist painter: HueValueChroma.com. All of you who have asked me
about color really need to visit this site and get this information into your artistic thought
processes. It's going to be a little rough going for some who shy away from technical language.
It's also going to challenge some of the conventional color "wisdom" that has been taught in
art schools for years. I personally find the information fascinating and VERY USEFUL ... I
hope HueValueChroma will give you more control over your color choices as it has me.

Douglas Flynt (USA), fine artist and painting teacher at the Grand Central Academy of Art,
New York:"Huevaluechroma.com" is a great resource to better understand color and how light
affects color.

Slade Wheeler (USA), fine artist. His site hosts a large amount of well organized/concise
information coupled with informative illustrations, including 3D modeling and animations, all
of which make this one of the best online color theory resources that I've been able to find.

Colour Research Society of Canada/Societe canadienne de recherche sur la

couleur (Canada). Excellent overview of the dimensions of colour and light perception for
painters & digital media artists;

Danny Pascale (USA), CEO of BabelColorR colour measurement and analysis: A well
illustrated site on light, color, and its perception. The content is a course in applied color
science optimized for artists but useful for all. The language is clear, with just a few simple
equations and lots of descriptions.

Mary-Angela Papalaskari (USA), lecturer, Department of Computing Sciences, Villanova

University, Pennsylvania. ... a set of webpages that give a great overview of color as it is
perceived, from the artist's perspective.

William Cromar (USA), artist, lecturer and Art Program Coordinator, Abington College, Penn
State University. "Color is a fascinating topic which we've only been able to scratch the surface
of in this title" [ART 314 - Material Culture: Light and Color]. "If you wish to go in greater
depth, visit David Briggs' comprehensive website The Dimensions of Colour."

Thomas Scholes (USA), digital artist and painting instructor, moderator

of Futurepoly forum: I think the best advice I can give is to study light in terms of physics, this
website is a great resource in those regards.

Josh Yavelberg, Professor, Art Institute of Washington (USA). The Dimensions of Color: an
in-depth guide on the concepts of the various color systems and how to move within each color
space by David Briggs. I urge you to go to the second, and onward, pages of each area as there
are interactive flash demonstrations of the various color spaces.

Chris Raadjes (UK), game artist at Auroch Digital Ltd. Probably the most intensely scientific
approach to light for painters [that's] available on the web. ... I'm struggling myself, but it's
worth it.

Daz Watford (UK), video game developer, concept artist: Now this website is big and
intimidating; but it's a great explanation of how colour created by light works. It's quite
sciencey and took me three goes to start to "get it", but it's worth the struggle. It will change
your understanding of colour with a "mind = blown" Inception ...

Paul Foxton (UK), fine artist, author of website Learning to See: ... this site has more
information than any site should really be allowed to have in one place. David's site is nothing
short of incredible. There's so much information there, and it bears such careful and close
reading, that I can only take it in bite sized chunks. I read half a page and have to think about
it for a week. This the best site about colour I know of. The relevance of all of it to painting may
not be apparent to you straight away, and it may appear too scientific for 'feeling' types. But I
find myself mulling over things I've read there as I work, and it always results in deeper
insights into the way we perceive light and colour. Very highly recommended.

ALISON online training (UK): This course is ideal for any learner who practices the visual
arts, either professionally or as a hobby, and who wants to greatly enhance their knowledge
and understanding of colour theory.

Atelier Art Classes, Brisbane (AUS): An incredible resource for the painter [and] a
fascinating and informative resource for anybody who has an interest in the perception of
colour.

Joe Collins, Draw Academy (USA):Â If you ever want to read up on [colors], David Briggs
gives the most complete treatment I've ever found, can't recommend that website enough.

Bjoern Gschwendtner, Classical Atelier @HOME (Germany): Read more about color
on huevaluechroma.com. This is THE RESOURCE for color in the internet for artists.

Michael Hosticka (USA), recent Game Art & Design graduate, Ringling College of Art and
Design: I learned more about practical application of color within 10 minutes of reading that
than I have in all of my art classes combined.... I would highly recommend the website to
anyone who wants to improve their understanding of light and color and doesn't mind
technical reading.

Unless otherwise indicated, all material on this website is copyright Dr David

Briggs, 2007-2015, and is licensed for personal and commercial use under the
terms of a Creative Commons Attribution-No Derivative Works 2.5 Australia license.
Interactive demonstration of additive colour mixing. Drag the top, middle and bottom
triangular yellow sliders to the left to control the brightnesses of the red, green
and blue spotlights respectively. Copyright David Briggs and Ray Kristanto,
2007.

Next: Colours in Space

PART 1. THE DIMENSIONS INTRODUCED

1.1 Colours in Space
 Hue, Lightness and Chroma
 Painting Colours in Space

Hue, Lightness and Chroma

All painters, whether working in traditional or digital media, are in a real sense
navigators in space. Whether they are aware of it or not, each touch of colour
they apply can be considered, using various systems, as a point within a space
defined by three dimensions.
Figure 1.1.1. Left: Portrait of Vincent Van Gogh by Henri Toulouse
Lautrec. Pastel, 1887. Right: RGB colours from this image, plotted in
YCbCr colour space using the program ColorSpace by Philippe
Colantoni. (www.couleur.org).

The three-dimensional system most familiar to painters is the classification of

colours of objects according to dimensions of hue, lightness (= value or
greyscale value) and chroma or relative chroma (often loosely referred to as
"saturation". Hue refers to the circular scale of "pure" or "saturated" colours
formed by the colours seen in the spectrum (red, orange, yellow, green, cyan,
blue and violet), together with the non-spectral colours like magenta, seen when
the two ends of the spectrum are mixed. Lightness refers to the scale from black
to white; tone, value, and greyscale value are synonyms or very closely
related. Chroma or colour strength refers to the amount of visual difference
from a grey of the same value. In short then, the system may be said to classify
any object colour according to the closest full or "saturated" colour, the closest
grey, and the visual difference from that grey (Fig. 1.1.2). Because the maximum
chroma attainable with any set of paints varies with value, a hue page that
shows all variations of value and absolute chroma attainable for a given hue has
an irregular right margin (Fig. 1.1.2), and because this margin varies for each
hue (Fig. 1.1.3), a hue-lightness-absolute chroma space has an irregular, tree-
like shape. Simpler systems arbitrarily treat the maximum chroma attainable for
each hue as uniform, resulting in a circular "colour wheel" and a cylindrical
colour space.

Figure 1.1.2. Explanation of the dimensions of hue, value/

lightness and chroma in the Munsell system. Grey scale and 10YR hue
page from the Munsell Book of Color, Glossy Edition.

Colour order systems based on hue, lightness and relative chroma first appeared
in the early 19th century, but the key concept of absolute chroma was devised by
the American artist and art teacher, Albert Munsell (1858-1918), and published
in a small book entitled A Color Notation (Munsell, 1905). Munsell published
quantitative scales of hue, value and chroma in an Atlas of physical colour
chips (Munsell, 1915), which after his death was elaborated by the Munsell Color
Company (directed by his son Alex) as The Munsell Book of Color (1929). This
was further refined and developed by the Optical Society of America,
culminating in the "renotation" published in 1943, which related the Munsell
System to the world standard system of colorimetry developed by
the International Commission on Illumination or CIE (the abbreviation is
based on the title in its French form, Commission Internationale de
L'Eclairage). This 1943 "renotation" forms the basis of all subsequent editions.
The Munsell Book of Color has 40 hue pages (Fig. 1.1.3), and is available in a
choice of editions having either matte or glossy colour chips. An alternative hue-
lightness-chroma system, CIE L*H*c space, is also available as a physical atlas
of colour chips, the RAL Design Atlas. Hue, lightness and chroma are included
by the CIE as three of the six defined attributes of perceived colour, of which the
Munsell and CIE L*H*c systems provide two alternative sets of quantitative
measures.
Figure 1.1.3. The forty hue pages of a modern edition of the Munsell Book
of Color, Glossy Edition. Click on each hue page to enlarge, and scroll
down for more pages. In the matte edition the range of colours tends to be
a little greater among the light colours and a little less among the dark
colours.

The dimensions of lightness and chroma apply specifically to colours perceived

as belonging to objects, as opposed to lights. Colours perceived as belonging
to lights (Fig. 1.1.4B) can be described in terms of three dimensions of hue,
brightness (perceived amount of light) and eithersaturation (colour purity, i.e.
perceived freedom from admixed white light) or colourfulness (colour strength,
a function of both brightness and saturation).
Figure 1.1.4. Attributes for colours perceived as belonging to (A) objects
and (B) light. Saturation refers to purity of colour of light, and can vary
throughout its range (white to monochromatic) at any level of brightness;
it is represented in B by the angle from the neutral axis. Colourfulness
refers to strength of colour of light, and can be thought of as saturation
times brightness; it is represented in B by the distance from the neutral
axis. Chroma (strength of colour of objects) depends on the colourfulness
(saturation and brightness) of the light given off by an object for a given
level of illumination. Chroma is necessarily zero at maximum and
minimum value (white and black respectively), and reaches its maximum
range at intermediate value levels.

Lightness and chroma apply to colours of objects seen in nature or depicted in

an image, as well as to colours of an image itself. This is true whether the image
surface reflects light (e.g. a photograph, painting, or projector screen), transmits
light (e.g. a stained glass window) or emits light (e.g. a computer monitor or TV
screen). However, areas of the visual field occupied by objects can also be seen
as light, and thus the dimensions of brightness, saturation, and colourfulness
apply not only to primary light sources but also to the light remitted by non-
luminous objects to our eyes.

Painting Colours in Space

Traditional colour theory discusses the "colour wheel" and the tonal scale, but
typically these are either not related to each other, or they are integrated in a
very simplistic way, as in the colour sphere of Johannes Itten. Many painters
thus spend much time making up elaborate paint mixing charts without
attempting to visualize the series of mixtures they generate as paths through
colour space, and so tend to rely colour "recipes" obtained by examining their
mixing charts to see how they mixed a particular colour previously. Lacking a
conscious three-dimensional conceptual framework for colour, many painters
vaguely think of colours being "warmer" or "cooler", without troubling to
consider what they mean in terms of the more precise attributes of hue and
chroma. Traditional colour theory typically offers little guidance on the
physical principles involved in creating effects of light and shade, which require
the framework of colour space for their full explanation, and instead relies on
crude and inaccurate formulae, such as "get the shadow colour by adding the
complementary colour" and so on. A hallmark of traditional colour theory is the
admonition "Don't use black!". The real problem is not the black paint, but the
painter's inability to visualize any unintended effect of adding black paint as an
easily corrected shift within colour space (Fig. 1.1.6).

Students who have previously been exposed only to traditional colour theory are
frequently astonished when they first learn to think consciously of their
colouring activities as maneuvering through a three-dimensional colour
space. A three-dimensional conception of colour assists painters by providing a
framework (1) for observing colour relationships, (2) for selecting and
mixing colours, and (3) for creating colour relationships from the
imagination.

1. As a framework for observing colour relationships.

Painters trained in the concept of colour space do not try to copy each colour in
their subject in isolation (the strategy of every beginner). Instead, they use the
concept of colour space as a frame of reference for grasping the relationship of
each colour to the totality of colours present. Tonal realist painters, for example,
typically observe colour relationships in the light from their subject, and then,
by a process of either conscious or unconscious translation, identify each
individual colour in terms of the hue, value and chroma of the paint
colour they will need to use in order that the whole ensemble replicates the
visual appearance of the subject as closely as possible. In practice, this usually
involves first selecting the most important ten or so colours in the subject, and
finding the place of these in relation to each other (Fig. 1.1.5). This begins the
process of building what I call a scaffolding for progressively finding the place
of all remaining colours, most of which can usually be considered as variations
on, or intermediates between, these scaffolding colours.
Figure 1.1.5. : Left: Lyndall by David Briggs, 2005, oil on canvas. Right: plan view
(above)
and side view (below) of ten selected colours from the image plotted in YCbCr
space using
the programme ColorSpace by Philippe Colantoni.

2. As a framework for selecting and mixing colour.

Artists who think in terms of colour space do not need to remember recipes for
mixing colours: they understand that most colours can be mixed from any
number of combinations of paints, as long as the target colour is within the
three-dimensional gamut of those paints. They literally visualize colour mixing
as moving colour from place to place through colour space. They decide on the
changes in hue, chroma and lightness required, and predict in advance what
effect various additions are are likely to have. These crafty painters can mix
every colour they want very quickly and accurately, particularly if they equip
their palette with a series of strings of pre-mixed pools of colours at various
values. This approach to colour mixing was developed to an elaborate degree by
the influential mid-20th century American teacher Frank Reilly, whose
approach has been described in books by his ex-students including Apollo
Dorian, Frank Covino, Jack Faragasso and Angelo John Grado.

Figure 1.1.6. When a mixture of a colourant such as Cadmium Red Deep

with white paint (A) is darkened using Ivory Black, the paint mixture
decreases rapidly in chroma (B), whereas the paint mixtures needed to
represent an object of colour A in shadow decrease less rapidly in chroma
(blue arrow). The required shadow colour (C) can be obtained by adding
just enough of the the original colourant (Cadmium Red Deep) to restore
the required amount of chroma. (Adding black may also result in a hue
shift, which can be rectified by adding a small amount of paint of an
appropriate hue and the same value).

3. As a framework for creating colour relationships from the

imagination.

The dimensions of colour form an essential conceptual framework for any kind
of activity that involves creating colour relationships from the imagination. In
the nineteenth and early twentieth centuries, much thought on colour spaces
(including Munsell's own writings) was directed towards discovering rules
of "colour harmony", and there are still many echoes of this kind of
investigation today. On this site however I am much more concerned with the
application of colour space to the creation of convincing effects of light. The
concept of colour space provides a quantitative framework for applying the
simple physical laws that govern the behaviour of light and colour, some of
which were understood in a qualitative way as far back as Leonardo. If the artist
gets these relationships right in a painting, the payoff is not merely technical
correctness but can be a vivid glow of light and feeling of atmosphere. And, just
as with, for example, perspective and anatomy, having the understanding that
allows you to do something from the imagination makes working from nature
far more efficient.

Figure 1.1.7. Imaginary sphere under three imaginary light sources,

painted as three layers in screen mode (one for each light source). David
Briggs, 2007, Photoshop CS2.

Many painters think of colour space in terms of relative hue, lightness and
chroma, but there others who train themselves to think in terms of absolute
scales of these dimensions such as those of the Munsell Book of Color
(q.v. Graydon Parrish and Steve Linberg's Classical Lab). The glossy version of
the Munsell "big book" is favoured over the matte version among oil painters
because paint mixtures can be tested on the individual removable colour chips
and then wiped off safely. Painters who stop short of going the full Munsell
often find it very helpful to at least think of lightness in terms of an absolute
scale of some kind.

In the remainder of this introductory section we will examine each of the major
dimensions or attributes of colours in turn, but in order to really understand our
subject we must first take up the thorny question of what these "colours"
actually are!
Modified February 19, 2017. Original text here.

<< 1 2 3 4 5 6 7 8 >>

1.2 The Dimensions of What, Exactly?

 What is a Colour?
 Perceived Colour and Psychophysical Colour
 Philosophical Theories of Colour

What is a Colour?

What gives a light a particular hue, brightness and saturation? How does an
object come to have a particular hue, lightness and chroma? We perceive lights
and objects to have particular colours, but what is not clear to introspection
alone is whether the attributes of these perceived colours, such as hues like red
and green, are (1) physical properties residing in lights and objects that our
visual system simply detects, or (2) ways of seeing certain properties of lights
and objects, ways our visual system creates. If the latter, we must make a
distinction between colour as a psychological perception and the properties of
lights and objects that we see as these perceived colours.

Figure 1.2.1. A. Extract from Newton's Opticks (1704, p.90). B. Newton's

colour circle from the Opticks (1704, Book I, Part II, Pl. 3), illustrating his
"center of gravity" principle for predicting the colour of a mixture of
lights. Unequal amounts of red, orange, yellow, green, blue, indigo and
violet "rays", indicated by the relative sizes of the small circles (p, q, r, s,
t, u, and x), mix to make light of colour z between spectral orange (Y) and
white (O), seen as whitish orange.

Scientific consensus is firmly with the second of these two views, which can be
traced back to antiquity via Descartes, Locke and Galileo, but which begins in
substantial detail with Sir Isaac Newton's researches into the physical basis of
colour. In the well-known passage from his Opticks (1704) shown in Fig.1.2.1A
Newton explicitly distinguishes between colour as a psychological perception
("sensation") and what we call colours "in the rays" and "in the object". For
Newton the colour of a light is its "power" or "disposition" to be seen as this or
that perceived colour. Newton demonstrated elsewhere in the Opticks that for
an isolated light this power or disposition depends on the relative balance of the
component "rays" (we would now say wavelengths) present, represented
approximately by their "center of gravity" in his colour circle (Fig. 1.2.1B). But as
his diagram implies, most colours of light can be evoked by many different
combinations of "rays" having the same centre of gravity. This means that a
given colour of light does not correspond to a single physical combination of
"rays" but to a whole class of combinations (we would now say spectral power
distributions, as in Fig. 1.2.3, left) that are indistinguishable to the human visual
system. We now call members of such a class metameric, and we believe that
they are indistinguishable to human vision because they evoke the same relative
response of the three cone cell types on which our colour vision depends.

Thus in the Newtonian view the particular colour of a light, specified by its
position in his circle, is neither a purely physical nor a purely psychological
property, but involves a relation between the two, now called psychophysical.
The particular colour of an object for Newton depends on the object's
"disposition to reflect this or that sort of rays more copiously than the rest" (we
would now say its spectral reflectance), but similarly a whole class of spectral
reflectances are indistinguishable (metameric) to the human visual system
under the same illumination, so the particular colours of objects are also
psychophysical.

Perceived Colour and Psychophysical Colour

The International Lighting Vocabulary of the International

Commission on Illumination (CIE)1, whose terminology is widely
accepted as standard in science and technology, defines two distinct
concepts of "colour":

Perceived colour: "characteristic of visual perception that can be described

by attributes of hue, brightness (or lightness) and colourfulness (or saturation
or chroma)" (CIE, 2011, 17-198).

Psychophysical colour:"specification of a colour stimulus in terms of

operationally defined values, such as 3 tristimulus values" (CIE, 2011, 17-197).
Tristimulus values are in turn defined as the "amounts of the 3 reference colour
stimuli, in a given trichromatic system, required to match the colour of the
stimulus considered", for example RGB and XYZ values in the CIE colorimetric
system (CIE, 2011, 17-1345).

Figure 1.2.2. Members of a psychophysical colour class match in colour to

a "standard observer" when presented in the same context, but can evoke
different perceived colour attributes when presented in different contexts,
even when they are physically identical. The central squares on either
side are physically identical and so have the same psychophysical colour -
they have the same RGB specification (R159 G170 B65) and would match
the same Munsell chip placed against them - but they differ in perceived
colour.

These definitions correspond to two distinct ways in which the word "colour" is
used in common speech as well as in science and technology: Demonstrations of
contrast phenomena such as Fig. 1.2.2 can been described either by saying that
the influence of the surround affects the colour of the squares, or by saying that
on different surrounds the squares appear to be different colours, but
are actually the same colour. In the first description the word colour is used for
an aspect of immediate visual perception (perceived colour), in the second it
refers to the shared colour specification (psychophysical colour) of the squares,
one of the 16.7 million "colours" available on an RGB screen. Applying CIE
terminology, the same psychophysical colour evokes two different perceived
colours.

The six attributes mentioned in the definition of perceived colour - hue,

brightness, lightness, colourfulness, saturation and chroma - are all defined
perceptually, that is, as attributes of visual perception rather than as physical
properties of lights or objects. None of the six are alternative names for the
same thing: a perceived colour can be described in terms of hue, lightness and
chroma if it is seen as belonging to an illuminated object, or in terms of
either hue, brightness and saturation or hue, brightness and colourfulness if it
is seen as belonging to light. Brightness is how we perceive the amount of light
emitted, transmitted or reflected by an area, while lightness is how we perceive
an object's diffuse luminous reflectance, that is, its efficiency as a diffuse
reflector of light. Hue, comprising perceptions in the cycle of red, yellow, green
and blue, is how we perceive the direction of bias of the spectral power
distribution of a light relative to daylight, or of the spectral reflectance of an
object relative to a spectrally even reflector. Saturation or freedom from
whitishness is how we perceive the relative amount of spectral bias of a light,
while colourfulness is how we perceive the absolute amount of spectral bias; it
is the cumulative effect of the saturation and brightness of a light. Chroma or
visual difference from a grey of the same value is how we perceive the absolute
amount of spectral bias of an object's reflectance/transmission, that is, its
efficiency as a spectrally selective reflector/transmitter of light.

We normally perceive the object colour attributes of hue, lightness and chroma
instantly and automatically, without the need for conscious judgement or
effort. This ability evidently relies on pre-conscious processes that arrive at a
probable interpretation of the information from our eyes in terms of object
colours and illumination. For example, they arrive at an interpretation of which
areas in the visual field remit little light because they are dimly lit, and which
remit little light because they are poor reflectors of light. So when we look at
most areas of the visual field we automatically see, not a patch of light, but a
combination of an object with a seemingly intrinsic colour (the colour the object
would appear in daylight) and an illumination with a seemingly intrinsic colour
(hue, brightness and saturation). For example, a white sheet of paper viewed
clearly under yellow illumination is usually perceived as being white, the
yellowishness being automatically and unconsciously attributed to the colour of
the illumination, and a white sheet of paper viewed clearly in dim illumination
is also usually perceived as being white, the darkness similarly being attributed
to the level of illumination.

Whatever their (highly disputed) nature, these pre-conscious processes are

effective enough to give the perceived colour of an object a high degree of
stability (object-colour constancy) under different levels and (to a point) under
different colours of illumination. The combination of effortlessness and stability
creates and maintains the illusion that these perceived colour attributes, which
are generated at a post-receptoral level, reside in the objects themselves, where
we directly and passively "detect" them. Though they seem to be detected
directly and passively, these mentally generated object colour attributes are
more like elements of a 3D computer model projected by our minds onto the
external world.
Figure 1.2.3. Left: Spectral power distributions of three standard
white illuminants (light sources). Though very different physically, these
three lights have the same chromaticity (psychophysical colour considered
separately from brightness/luminance) and appear the same colour (a
warm white). Right: CIE xy chromaticity diagram showing chromaticities
of the RGB lights of a standard RGB colour space (sRGB) and of the
three illuminants shown on the left. The faint horseshoe-shaped line
shows the chromaticities of the monochromatic wavelengths of the
spectrum. All images derived from the program ColorSpace by Philippe
Colantoni.

A psychophysical colour specification identifies a class of physically varied

lights or objects that match in colour (are metameric) when viewed in the same
context. (Members of a psychophysical colour class can evoke different
perceived colours when presented in different contexts, even when they are
physically identical, as in Fig. 1.2.2). As "normal" colour vision varies somewhat
between individuals, psychophysical specifications use the necessary
assumption of a mathematically defined "standard" human observer. Every
metameric class of lights can be specified by the quantities of three standard
red, geen and violet lights, the CIE RGB "primaries", that match it, though this
specification involves a negative quantity whenever one of the primaries must
be added to the test stimulus to obtain a match. CIE RGB values are commonly
converted to CIE XYZ values in which the Y is equal to the visible energy of
light, called luminance and perceived as brightness.

A three-dimensional set of CIE RGB or XYZ values specifies the psychophysical

colour of a light including its brightness/ luminance, just as a set of familiar
RGB values does for a digital colour. We can also specify the colour of a light
considered separately from its brightness/ luminance, called its chromaticity,
using the ratio of the CIE X,Y and Z components, and represent this on a two-
dimensional triangular CIE xy chromaticity diagram (Fig. 1.2.3).
This diagram is a descendant of Newton's colour circle, in which the spectrum
now forms a horseshoe shape instead of a circle, but in which the colour of a
light still depends on the balance of its spectral components. Despite their very
different spectral power distributions the three CIE illuminants (light sources)
shown in Fig. 1.2.3 have the same chromaticity, and all appear "warm white" to
an observer with normal colour vision. The CIE xy diagram may be familiar to
photographers and digital painters from its use to compare the gamuts (colour
ranges) of different colour spaces or of the actual RGB lights used in different
models of monitor.

Colours of objects can be specified psychophysically by the CIE RGB or XYZ

tristimulus values of the light they reflect/ transmit, but as these values depend
on the illuminant used it is necessary to also specify the latter, normally a
standard "white" daylight illuminant such as CIE D65 or the older CIE
Illuminant C.

Figure 1.2.4. Identical range of digital colours (sRGB colour space)

represented in two different psychometric colour spaces. A. Munsell
colour space, specifying colours in terms of Munsell hue, Munsell value
and Munsell chroma. B. CIE L*a*b* colour space, specifying colours in
terms of CIE lightness (L*) and two orthogonal chromatic dimensions
corresponding to reddish vs greenish chroma (a* vs -a*) and
yellowish vs bluish chroma (b* vs -b*). These chromatic dimensions can
alternatively be presented as measures of hue H* and total chroma c in
the form of CIE L*H*c space. Illustrations prepared using drop2color by
Zsolt Kovacs-Vajna (A) and ColorSpace by Phillippe Colantoni. (B)
As attributes of perception, hue, brightness, lightness, colourfulness, saturation
and chroma can not be measured directly, but various
quantitative psychometric scales have been devised against which these
psychological attributes can be judged. These psychometric scales can be
presented purely as a collection of coloured samples, as in the 1915
Munsell Atlas and the 1929 Munsell Book of Color, but modern psychometric
scales involve either an atlas of samples tied in with a table of colorimetric
equivalents, as in the modern Munsell Book of Color and the
Scandinavian Natural Colour System (NCS) atlas, or they are based directly on
colorimetric data, as in CIE L*a*b* space. The tristimulus values for an object
can thus be converted via a table or a formula to a position in a psychometric
colour space that prediucts the perceived colour of that object under specified
standard conditions. Since all major psychometric object colour spaces specify
perceived colour under some form of daylight illuminant, it could be said that
these specifications (e.g. Munsell hue, value and chroma) describe the
seemingly intrinsic colour of an object that object-colour constancy to varying
degrees recovers.

The alternative psychometric measures used in different systems for the same
attribute typically agree in general terms in the way they arrange psychophysical
colours, but differ in fine detail (Fig. 1.2.4). In the following pages we will
introduce each of the six colour attributes defined by the CIE plus some
alternative attributes defined in other systems including the Scandinavian NCS,
along with various psychometric scales used as quantitative measures of some of
these attributes.

Philosophical Theories of Colour

Colour has attracted a great deal of attention from philosophers in recent

decades because of its connections with major metaphysical issues including the
nature of physical reality, perception and the mind. Unfortunately a substantial
part of the discussion has been taken up with what appear to amount to
semantic disputes over the one true referent of the word "colour". For
example, eliminativism confines the term "colour" to perceived colour, and thus
concludes that colours are illusory in the sense that they are perceiver- and
species-dependant, while dispositionalism confines "colour" to the
(psychophysical) disposition of lights and objects to cause an experience of
perceived colour. These theories isolate different parts of the scientific
consensus, from Descartes and Newton to the CIE, that we use the word
"colour" in two senses, for a psychological perception and for the properties that
we see as those perceptions. More eccentrically, colour physicalism equates the
colour of an object with its particular spectral reflectance, classifying its
"perceived colour" as merely the appearance of this actual, physical colour. Most
of the major philosophical positions seem more or less compatible with current
science on the process of colour vision, but colour primitivism on the other
hand valiantly defends the position that colours are intrinsic, non-relational,
and non-reducible properties of lights and objects. For an infinitely fairer and
more authoritative outline of the philosophy of colour than I can provide, the
reader should consult Barry Maund's entry "Color" in The Stanford
Encyclopedia of Philosophy, written from the perspective of one of the leading
exponents of the field.
1
The International Commission on Illumination or Commission Internationale de L'Eclairage
(CIE) was founded in 1913 and is the organization responsible for the international coordination
of lighting-related technical standards. In the field of colour the CIE established the fundamental
framework of modern colorimetry including the CIE 2o and 10o standard observers, the CIE
standard illuminants (Illuminants C, D50, D65 etc), various widely used colour spaces including
CIE XYZ, CIE xyY and CIE L*a*b*, colour difference formulae (CIEDE94, CIEDE2000), and colour
appearance models including CIECAM02. Its International Lighting Vocabulary lists definitions
for 1448 terms and is by far the most comprehensive and authoritative source on the scientific
terminology of light and colour.

Page added January 15, 2017

<< 1 2 3 4 5 6 7 8 >>

1.3 The Dimensions of Colour: Lightness

 Lightness, Value and tone
 Measuring Lightness
 Lightness in Digital Colour Spaces

Lightness, Value and Tone

Figure 1.3.1. A. Sunlight Sweet, Coogee (1890) by Arthur Streeton. B,

value/ lightness and C, hue and chroma information.
Lightness, value, greyscale value, and tone are all very closely related or
identical in meaning. Value and greyscale value refer to the scale from black
through various shades of grey to white, increasing in that order. Lightness as
officially defined by the CIE it applies to light-transmitting as well as light-
reflecting objects, though in Arend's (1993) "slightly simplified and clarified"
definition as "apparent reflectance" it is confined to light-reflecting objects and
is in this usage is an exact synonym of value. The word tone as used by painters
refers (in one of its senses) to the same scale, but is sometimes deemed to
increase from white to black.

Lightness: "brightness of an area judged relative to the brightness of a

similarly illuminated area that appears to be white or highly
transmitting" (CIE 2011, 17-680).

Lightness is how we perceive the physical property of diffuse

luminous reflectance of an object, that is, it is our perception of its efficiency as
a diffuse reflector of light. Of all of the wavelengths physically reflected by an
object, lightness is influenced only by those wavelengths visible to the human
visual system (that is, light), to the extent of their influence on that system,
which is highest near the middle and lowest at the ends of the spectrum.

Lightness does not apply to an area perceived as a primary light source, because
primary light sources have the potential to increase in brightness indefinitely,
and so their colours typically are not judged in relation to an upper limit
perceived as white. Primary light sources can thus be said to be brighter or
dimmer, but not lighter or darker grey. However colours on luminous computer
and TV screens are seen as having lightness if the screen is perceived, as it
normally is, as a coloured object instead of as a primary light source. Here the
finite range of brightnesses of the pixels at a given brightness setting of the
display create a context within which llightness is perceived.
Figure 1.3.2. Tonal sketches of two of Howard Pyle's illustrations from
Andrew Loomis' Creative Illustration (1948).

In artworks and in perception of images in general, lightness is the pre-eminent

attribute or dimension of colour. If the lightness information in an image is
isolated from the chromatic information, remarkably large amounts of the
legibility and compositional structure are preserved (Fig. 1.3.1). In traditional
art and design education, lightness (as value or tone) is often treated separately
from and earlier than chromatic colour as one of the fundamental elements of
composition, alongside line, shape, mass, rhythm etc. Manipulating the
lightness structure of a picture by "massing" values into a striking arrangement
of shapes, and choosing the distribution of lightnesses through the scale to
establish an expressively appropriate value "key", are well known traditional
compositional strategies (Fig. 1.3.2).

Figure 1.3.3. Cleland's (1921) illustration of Munsell's conception of his

system as a tree of colours.

In an early exposition of the Munsell system Cleland (1921) expressed the

central importance of value with an image of a "Color Tree" (Fig. 1.3.3). For the
painter, value is to colour as tree trunks are to trees.

Measuring Lightness
Figure 1.3.4. CIE 1931 luminous efficiency function, showing the relative
response of the human visual system to light at each wavelength of the
spectrum. The luminance or visible energy of a light can be calculated by
weighting its spectral power distribution with this function, or measured
directly using a photometer equipped with a filter that transmits each
wavelength in these proportions. Curve from http://www.cvrl.org/, which
also illustrates various (subtle) refinements of the 1931 function that have
been subsequently proposed.

Psychometric measures of lightnerss are based on the ratio of the luminance of

an object to the luminance of a reference white under the same illumination,
where luminance (Y) refers to visible energy of light, that is, light energy
weighted according to the wavelength-by-wavelength response of the human
visual system (Fig.1.3.4). This luminance ratio is known as the diffuse luminous
relectance of the object and ranges from a few percent for glossy black oil paints
to around 90% for titanium white oil paint and up to 99% for a matte
magnesium oxide coating. Munsell value is calculated based on luminance
relative a theoretical perfect white reflector, which is assigned a Munsell value of
10, while CIE lightness L* on the other hand is calculated based on luminance
relative to an actual similarly illuminated standard white reflector, which is
assigned a lightness of 100. Munsell value 10 consequently has a CIE lightness
slightly greater than 100 (Fig. 1.3.5B).
Figure 1.3.5. A, Munsell value scale for painting classes (David Briggs,
2012). B. Comparison of Munsell value with CIE lightness (L*). C, Value
scale of Denman Ross, from Snow and Froelich (1904).

However, both Munsell value and CIE lightness have a nonlinear relationship to
these luminance ratios, such that a middle grey object on these scales has only
about 20% of the luminance of a similarly illuminated white object. CIE
lightness uses a different nonlinear transformation to the Munsell system,
giving numbers roughly but not exactly ten times the Munsell value of the same
grey. This nonlinearity is introduced in order to achieve approximate perceptual
equality of steps. as it takes a greater absolute increase in luminance to create
the same amount of contrast at high lightness than at low lightness. A object
reflecting 20% of the light energy of a white object appears to be about as many
steps away from black as from white, but it should not be inferred from this that
human perception of relative luminance is nonlinear. If you ask yourself how
much light appears to be coming from a middle grey area compared to a nearby
white area, you will probably estimate something like 20% rather than 50%.

In Albert Munsell's original value scale of 0 to 10, actual black and white paints
had values of 1 and 9 respectively, giving the same number of steps as in the
slightly earlier system by Denman Ross (Fig. 1.3.5C). Some painters working in
traditional media today continue to use an informal scale of five or nine evenly
spaced values between black and white inclusive. In the modern ("renotated")
Munsell system, black and white glossy oil paints attain values of about 0.5 and
9.5 respectively, leaving nine intermediate values for coloured paints.

Figure 1.3.6. Helmholtz-Kohlrausch Effect. A. Various digital colours on a

grey background, all measuring L = 50 in Adobe Photoshop, and thus in
theory of equal luminance. The strongly coloured areas show a kind of
glow that can make them seem brighter and lighter than the
equiluminant grey. Actual equiluminance depends on the display settings,
but may be obtained by varying the angle of view. For each colour vary
the angle of view until it seems equiluminant to the grey background
(judged by squinting) then open you eyes to see the effect in the strongly
coloured areas. B, same image converted to greyscale mode.

The lightness of any object colour is the lightness of the grey it most closely
resembles. A substantial ambiguity of the term arises from the fact that strongly
coloured objects are perceived to have a certain chromatic "glow" or brilliance
that can create the impression that they are lighter than the grey that would be
measured to be equal in luminance under similar lighting. This equiluminant
grey would nevertheless show the least contrast with the object at their border
when placed in contact. In some colour appearance models this colour glow,
called the Helmholtz-Kohlrausch effect, is included as a legitimate component
of an area's lightness and brightness, whereas in colour order systems such as
the Munsell system and CIE L*a*b* it is not, and the criteria of least-contrast
and equality of luminance apply in judging value. The painter's trick
of squinting (viewing through the eyelashes of nearly closed eyes) diminishes
this colour glow and is extremely useful for comparing luminance/ CIE
lightness/ Munsell value. An excellent way to learn to recognize equality of
value/ luminance in colours of different chroma is by doing the hue page
exercise included in the New Munsell Student Color Set. Some online
versions of this exercise have been created using Flash by Orian Lima.

Lightness in Digital Colour Spaces

Lab space, used in the colour picker and as an image mode in Adobe Photoshop,
is closely based on CIE L*a*b* colour space, and its lightness dimension L is a
reliable measure of value on a scale of 0 to 100. Any adjustment in which it is
desired to keep lightness constant or to change it in a predictable way is best
done in Lab mode.

Figure 1.3.7. Relationship of HSB "brightness" B to Lab lightness L.

Inset: external and cross-sectional views of HSB colour space represented
as a cone, the latter showing lines of equal HSB "brightness" B
(horizontal) and HSB saturation S (radiating from black).

The parameter of so-called "brightness" B in HSB colour space, also known as

"value" V in the alternative name of this space, HSV, is neither absolute
brightness nor lightness, but is a measure of brightness or lightness relative to
the maximum possible for RGB colours of a given HSB hue angle H and
saturation S. RGB colours with an HSB "brightness" of 100 range from 30 to
100 in Lab lightness (Fig. 1.3.7).

Figure 1.3.8. Relationship of HLS "lightness" to Lab lightness. Inset:

external and cross-sectional views of HLS colour space, the latter showing
lines of equal HLS "lightness" L (horizontal) and so-called saturation S.

HLS is another computationally simple colour space devised in the early years
of digital graphics as a more intuitive alternative to RGB for choosing and
manipulating colours. It is still used in many graphics programs, including the
"Hue/Saturation" adjustment panel and the "Desaturate" command in Adobe
Photoshop (RGB mode only). Its dimension of so-called "lightness" L has a very
tenuous relationship to perceived lightness. Colours that have an HLS L of 50
include the highest-chroma colour of each hue and all colours directly
intermediate between these colours and middle grey. The resulting colour space
has the form of a symmetrical double cone in which colours on the L=50 plane
vary from 30 to 98 in Lab lightness (Fig. 1.3.8). When applied in Photoshop in
RGB mode, the "Desaturate" command converts all colours to the grey having
the same L in HLS space, with the result that all highest-chroma colours convert
to the same middle grey. The same command when applied in Lab mode
converts all colours to the grey having the same L in Lab space, resulting in a
much more natural greyscale conversion.

Page added January 16, 2017. A much earlier (2007) account of lightness is
located on this site here

<< 12 3 4 5 6 7 8 >>

1.4 The Dimensions of Colour: Hue

 The Circuit of Hues
 Measuring Hue
 Subdividing the Hue Page

The Circuit of Hues

Figure 1.4.1. Ewald Hering's analysis of the hue circle into red/green and
yellow/blue components, from Outlines of a Theory of the Light
Sense (1964), an English translation of his posthumous Grundzuge der
Lehre vom Lichtsinn (1920). Hering's colour-opponent model was adopted
very early in psychology, but was not generally accepted in vision science
until the mid 20th century, when models showing it could be reconciled
with the trichromatic input at the retinal level gained widespread
support.

The hue of any colour is its closest match in the circuit of "pure" or "saturated"
colours known to artists as the colour wheel. Thus the hue of a brown object is
the particular orange-yellow to orange-red that it most closely resembles.

A perennial question from students in classes and on internet forums runs along
the lines of "why do we bend the linear scale of colours in the spectrum into a
circular colour wheel?". The assumption is that hues reside in the linear
sequence of wavelengths of the spectrum and must be bent into a circle, but
according to the widely held model of colour vision called colour opponency,
proposed in the late 19th century by Ewald Hering, hues are produced by the
visual system with an intrinsically circular range, only part of which can be
evoked by single wavelengths of light. But before the recognition of colour
opponency the question was indeed a baffling one, and the best explanation that
Newton could offer was his rather optimistic suggestion of an analogy with a
circle of musical notes.

The scientific definition by the CIE attributes the cycle of hues to successive
combinations of four hues identified in a separate definition as the "unique" or
"unitary" hues in acknowledgment of the concept of colour opponency. Colour
opponency is also explicitly acknowledged in the hue circle of the
Scandinavian Natural Colour System (NCS).

Hue: "attribute of a visual perception according to which an area appears to

be similar to one of the colours: red, yellow, green, and blue, or to a
combination of adjacent pairs of these colours considered in a closed
ring" (CIE, 2011 17-542).

Unique hue: "hue that cannot be further described by the use of hue names
other than its own. Equivalent term: "unitary hue". NOTE There are 4 unique
hues: red, green, yellow and blue forming 2 pairs of opponent hues: red and
green, yellow and blue." (CIE, 2011, 17-1373).
Figure 1.4.2. A. Normalized responses of human S, M and L cone types to
different wavelengths of light. Note that the M and L cones do not
"detect" green and red wavelengths respectively, as is often claimed in
popularized explanations. They both respond to photons through
virtually the entire spectrum, and if they are hit by a photon they can not
distinguish what part of the spectrum it came from, they simply respond
more strongly to photons from some parts than others. B. The mixture of
wavelengths in daylight stimulates all three cone types strongly, and the
light is seen as colourless (white), but any single wavelength must result in
unequal cone responses, which are experienced as different hues.

Hue is how we perceive the direction of bias of the spectral power

distribution of a light (its distribution of energy through the spectrum) or of
the spectral reflectance of an object (its reflectance through the spectrum).
Conversely, absence bias in either of these is perceived as absence of hue.
Perception of a circle of hues is specific to organisms whose colour vision
involves three receptor (cone cell) types processed in a cone-opponent fashion,
that is, by comparing the responses of the three cone cell types with each other
at an early stage of visual processing. When the energy of an isolated light is
distributed throughout the spectrum in a balance similar to that of daylight, the
responses of our long-, medium- and short-wavelength (L, M and S) cone cell
types are mutually balanced and cancel each other out chromatically, and the
light is perceived as colourless , that is, "white" (Fig. 14.4.2A)1. Single
wavelengths on the other hand necessarily evoke an unequal response from the
cone cell types (Fig. 1.4.2B). Beginning at the short wavelength end these
unequal cone responses are successively S, S+M, M, M+L and L dominant. An
L+S dominant response, which completes the cycle of possible cone responses,
can be evoked only by lights containing mixed wavelengths predominantly from
the two ends of the spectrum.

Figure 1.4.3. (A) yellow/blue and (B) red/green colour-opponent hue

perceptions evoked by different wavelengths, based on the hue
cancellation experiments of Hurvich and Jameson (see Fig. 3.3, Section
3.2).

We perceive this cycle of possible cone responses as a cycle of hues comprising

successive combinations red, yellow, green and blue, apparently organized as a
yellow or blue and a red or green colour-opponent pairs (Fig. 1.4.3A,B). It
remains controversial how these colour-opponent perceptions (or "signals")
arise from the circuit of trichromat cone-opponent responses (Hardin, 2014),
but in general terms a yellow vs blue perception is evoked by long or
middle vs short wavelength dominance, while a red vs green perception is
evoked by long OR short vs middle wavelength dominance (Fig. 1.4.3B). Note
that by this model, "red" is not a property of light that our visual system detects,
it is a perception that our visual system creates in response to two distinct bands
of wavelengths that have nothing in common except that they occupy opposite
ends of the range visible to humans. For a dichromatic species having only two
cone types, say L and S cones, the range of relative responses is confined to just
L dominant and S dominant, which can be perceived as just two hues. Human
dichromats are believed to perceive colour something like this, though there is
evidence that colour perception in dichromatic-testing individuals of a
trichromatic species can be more complex.

Figure 1.4.4. Spectral hues explained as successive combinations of

yellow/blue and red/green colour-opponent hue perceptions.

The cone responses evoked by single wavelengths (S, S+M, M, M+L and L
dominant) are experienced as successive combinations of these colour-
opponent pairs, forming the the spectral hues spectral red, orange, yellow,
green, cyan, blue, and violet (Fig. 1.4.4A,B). Middle ("pure") yellow and blue are
experienced where the red/green perception is at zero, and middle green is
experienced where the yellow/blue perception is at zero. The non-spectral
S+L dominant response, requiring mixtures of wavelengths from the two ends
of the spectrum, is seen as the range of non-spectral hues from purple to
magenta and middle red. Broadband lights containing a mixture of wavelengths
evoke a cone response and perceived hue that depends on the overall balance of
wavelengths present.

Figure 1.4.5.Comparison of spectral reflectance curves of

Gamblin Cadmium Yellow Light, Cadmium Yellow Medium and Titanium
White oil paint. Munsell plot and spectral reflectance curves for Gamblin
Conservation Colours from drop2color by Zsolt Kovacs Vajna; paint
sample photos from dickblick.com.

The colours we see as belonging to objects at least to a degree relate to their

characteristic spectral reflectance, rather than to the wavelengths that they
happen to reflect under a particular illuminant, and the hues of these object
colours depend on the the direction of bias of these spectral reflectances. Paints
that strongly reflect the middle and long wavelength parts of the spectrum (Fig.
1.4.5) are seen as strongly yellow (Fig. 1.4.3A), and either slightly greenish,
slightly reddish, or neither depending on the relative size of the middle and long
wavelength contributions (Figure 1.4.3B). In Gamblin Cadmium Yellow Light
these contributions are approximately evenly balanced and the paint is seen as
middle yellow, neither greenish nor reddish, while in Gamblin Cadmium Yellow
Medium the contribution from the middle (green-evoking) part of the spectrum
is smaller, and so the paint colour is seen as having a red component (Fig. 1.4.5).
Titanium White is seen as being white because its spectral reflectance has no
substantial bias.

Despite a broadly based scientific consensus to the contrary, the view that hues
like red and green are properties or "things" physically residing in objects is still
alive and well in traditional colour theory accounts of "colour mixing", which
take the seemingly logical additional step of assuming that when we mix paints
the hues residing in the paints themselves mix. The traditional doctrine that red,
yellow and blue are primary colours that "can't be mixed from other colours"
proceeds from these assumptions: we can't make a red mixture without using
paint that already "contains red" (such as magenta), similarly we can't make a
yellow or blue mixture without using paints that already "contain" yellow and
blue respectively, but the colour green is not a primary colour but a mixture of
colours because we can make a green mixture from middle yellow and middle
blue paints that don't "contain" green. (For the scientific alternative to this
impeccable phenomenology see Section 5:Subtractive Mixing).

Unfortunately, many popularized explanations of colour vision also encourage

the idea that hues exist outside us by speaking of cone cell types that "detect red,
green and blue wavelengths" respectively (which is absolutely incorrect, See Fig.
1.4.2A), especially when they take this to mean that humans really only see red,
green and blue, that mixtures of red and green lights just "make you think
you're seeing yellow", or that the colour magenta/pink/purple alone is "made
up" by our brain, because it doesn't "exist" in the spectrum! Although the
colour-opponent model has been widely accepted alongside trichromacy in
vision science and psychology for many decades, most popularized explanations
of colour vision still omit it completely and explain only the trichromatic model
at the level of the retina.

Measuring Hue

Figure 1.4.6. The forty hue pages of the Munsell Book of Color (glossy edition).

Hue systems differ among themselves by the criterion used for placing hues
opposite each other, which affects the spacing of the hues, and by which hues
that are treated as "primary" or principal in anchoring the circular scale. More
incidental differences are the sequential direction of the hues (i.e. clockwise or
counterclockwise spectral sequence), whether the continuous loop of hues is
represented the form of a circle, hexagon, triangle, star, or other closed
geometrical figure, and the conventional orientation (i.e. a specific hue placed at
the top of the figure or treated as the first in the sequence). Hue systems are
classified on this site according to the criterion used to place hues opposite each
other, and hue systems based on the historical primaries and their
complementaries, opponent hues, additive complementaries, colourant-mixing
complementaries and perceptually-equal spacing respectively are discussed in
more detail on later pages.

Figure 1.4.7. A. Hue circle based on the historical primaries red, yellow
and blue from The Art of Color (1961) by Johannes Itten. The graphic
visually encapsulates Itten's view that his secondary colours green,
orange and violet each "contain" two of the primaries. B. Hue circle
based on opponent hue relationships from the Natural Colour System
(NCS) by the Scandinavian Colour Institute.

Hue systems based on the historical primaries are organized around the
conventional complementary pairs red-green, orange-blue and either yellow-
purple or yellow-violet. They include numerous 18th-21st century examples of
the "artists' color wheel" of traditional colour theory. These hue systems
reflect historical confusion of colourant-mixing and opponent hue relationships.

In traditional colour theory the concept of warm and cool colours is commonly
used to label distinctions of hue, for example warm yellow for reddish yellow
and cool yellow for greenish yellow. These warm/cool associations could be
considered a mild but widespread form of synesthesia. As is typical of that
condition, individuals who perceive warm/cool associations of colours can be
adamant that the association they perceive is objective and obvious, yet other
individuals can perceive the precisely opposite association. A striking example
concerns the colour blue, where different camps within traditional colour theory
regard a reddish blue like ultramarine as a warm and a cool blue respectively.
Even if one feels strong warm/cool associations of colours, in the interests of
clear communication it is wiser to speak of reddish or greenish blue rather than
warm or cool blue, and so on.

Hue systems based on opponent hue relationships are organized around

the opponent hue pairsred-green and yellow-blue, and include some historical
examples and the modern Natural Colour System or NCS (Fig. 1.4.7B). Different
individuals show a rather wide range of variation in the colour chips they select
as representative of the four opponent hues, but the averages for most studies
show a reasonable range of 2 or 3 Munsell pages (Fig. 1.4.8). These studies show
that the average determinations for the unique hues are not perceptually equally
spaced.
Figure 1.4.8. Munsell hue-chroma plane showing colours for sixteen
common artists' pigments as pure paints and as 1:1 mixtures with
Titanium white paint, calculated using Zsolt Kovacs-Vajna's
program drop2color. The coloured circles around the circumference show
average determinations of positions of unique red, yellow, green and blue
from four studies using Munsell chips cited by Kuehni (2012); studies
involving conversion from spectral data, NCS hues etc show further
spread.

Hue systems based on perceptually uniform spacing oppose hues that are
separated by an equal number of perceptual steps in either direction around the
hue circle. They include the Munsell system and CIE L*a*b* system. Broadly
speaking, opposite hue pairs in these systems are close to additive complements,
especially compared to other systems. The Munsell hue circle (Figs 1.4.6, 1.4.8)
is based on five principal hues, red, yellow, green, blue and purple (R, Y, G,
B and P) and five intermediate hues (YR, GY, BG, PB and RP). Each of these ten
hues was originally intended by Albert Munsell to have ten numbered divisions,
with the fifth division, for example 5R, being regarded as the typical version of
the hue. The Munsell Book of Color however has always had only four hue pages
for each principal and intermediate hue, for example 2.5R, 5R, 7.5R and 10R,
making a total of 40 hues.

Figure 1.4.9. Hue circle showing hue angle (H) used to specify hue in
HSB, HLS and HSI digital colour spaces. The circle arranges the additive
complementary pairs digital red-cyan (0o-180o) , yellow-blue (60o - 240o)
and green magenta (120o-300o) opposite each other. Note however that
opposite colours are not complementary away from these three axes. The
hue spacing is far from perceptually even, as can be determined both by
visual inspection and by the approximate positions of the Munsell
principal and intermediate hues (source).
Hue systems based on additive complements oppose hues of lights that make
white light when mixed. They include systems by Newton, Helmholtz, Rood and
Ostwald, and the CIE L*u*v* system. Examples of additive complementary pairs
are familiar to many from the major hue axes in the HSB hue system used in
graphics programs: the digital hues going by the names
of magenta/green, red/cyan, and yellow/blue. HSB hue angle (H) is calculated
from the ratios of the (nonlinear) RGB primaries by a simple formula, and is
highly uneven perceptually (Fig. 1.4.9).

Figure 1.4.10. Colour circle diagrams by Michel Jacobs from (A) The Art
of Colour (1923) and (B) The Study of Colour (1925).

Hue systems based on colourant-mixing complementaries oppose hues of

paints or dyes that make a neutral grey or black when mixed. The regular
succession of hues in the traditional artists colour wheel (e.g. Fig. 1.4.6A)
promotes the assumption that paint-mixing complementary pairs succeed each
other in an orderly succession of straight lines forming diameters of the hue
circle. Many painters evidently paint for years without questioning this
assumption, but in reality mixing paths show markedly different patterns for
paints that are close to the ideal subtractive primary hues compared to paints
that are far from these hues. Paints close to orange-red, yellow-green and blue-
violet in hue each neutralize or nearly neutralize paints of a remarkably large
spread of opposing hues, so that to make a hue circle that places paint-mixing
complementaries approximately opposite each other it is necessary to expand
the sectors occupied by these hues and contract the ranges of the remaining
hues. Michel Jacobs used a diagram of precisely this sort (Fig. 1.4.10) in his once
popular books The Art of Colour (1923) and The Study of Colour (1925), both of
which were republished in numerous editions until 1956.
Subdividing the Hue Page

Figure 1.4.11. Four approaches to dividing up a hue page.

In three-dimensional colour systems each hue corresponds to a two-

dimensional array of colours that can be called a hue page. Hue pages in
different systems, such as Munsell, NCS and digital hue angle H, typically drift
through each other to some extent. We've seen that hue pages in the Munsell
system are divided up according to value and chroma, but the same array of
colours can be classified in many alternative ways for different purposes (Fig.
1.4.11). Thus the digital colour space HSB divides up a hue page according to
measures of saturation and relative brightness (Fig. 1.4.11B), while the digital
colour space HLS divides up a hue page according to rather arbitrary measures
of relative lightness and relative chroma, confusingly called "lightness" and
"saturation" (Fig. 1.4.10D). Several important systems including the historical
Ostwald system and the modern Scandinavian NCS divide up a hue page (in
different ways) into white, black and coloured components (Fig. 1.4.11C).
1
These comments refer to light viewed in isolation; the perceived colour of a light or object can
of course vary considerably depending on surrounding, interspersed and recently
viewed colours.

Page added January 21, 2017.

<< 12 3 4 5 6 7 8 >>

1.4 The Dimensions of Colour: Hue

  The Circuit of Hues
 Measuring Hue
 Subdividing the Hue Page

The Circuit of Hues

Figure 1.4.1. Ewald Hering's analysis of the hue circle into red/green and
yellow/blue components, from Outlines of a Theory of the Light
Sense (1964), an English translation of his posthumous Grundzuge der
Lehre vom Lichtsinn (1920). Hering's colour-opponent model was adopted
very early in psychology, but was not generally accepted in vision science
until the mid 20th century, when models showing it could be reconciled
with the trichromatic input at the retinal level gained widespread
support.

The hue of any colour is its closest match in the circuit of "pure" or "saturated"
colours known to artists as the colour wheel. Thus the hue of a brown object is
the particular orange-yellow to orange-red that it most closely resembles.

A perennial question from students in classes and on internet forums runs along
the lines of "why do we bend the linear scale of colours in the spectrum into a
circular colour wheel?". The assumption is that hues reside in the linear
sequence of wavelengths of the spectrum and must be bent into a circle, but
according to the widely held model of colour vision called colour opponency,
proposed in the late 19th century by Ewald Hering, hues are produced by the
visual system with an intrinsically circular range, only part of which can be
evoked by single wavelengths of light. But before the recognition of colour
opponency the question was indeed a baffling one, and the best explanation that
Newton could offer was his rather optimistic suggestion of an analogy with a
circle of musical notes.

The scientific definition by the CIE attributes the cycle of hues to successive
combinations of four hues identified in a separate definition as the "unique" or
"unitary" hues in acknowledgment of the concept of colour opponency. Colour
opponency is also explicitly acknowledged in the hue circle of the
Scandinavian Natural Colour System (NCS).

Hue: "attribute of a visual perception according to which an area appears to

be similar to one of the colours: red, yellow, green, and blue, or to a
combination of adjacent pairs of these colours considered in a closed
ring" (CIE, 2011 17-542).

Unique hue: "hue that cannot be further described by the use of hue names
other than its own. Equivalent term: "unitary hue". NOTE There are 4 unique
hues: red, green, yellow and blue forming 2 pairs of opponent hues: red and
green, yellow and blue." (CIE, 2011, 17-1373).

Figure 1.4.2. A. Normalized responses of human S, M and L cone types to

different wavelengths of light. Note that the M and L cones do not
"detect" green and red wavelengths respectively, as is often claimed in
popularized explanations. They both respond to photons through
virtually the entire spectrum, and if they are hit by a photon they can not
distinguish what part of the spectrum it came from, they simply respond
more strongly to photons from some parts than others. B. The mixture of
wavelengths in daylight stimulates all three cone types strongly, and the
light is seen as colourless (white), but any single wavelength must result in
unequal cone responses, which are experienced as different hues.
Hue is how we perceive the direction of bias of the spectral power
distribution of a light (its distribution of energy through the spectrum) or of
the spectral reflectance of an object (its reflectance through the spectrum).
Conversely, absence bias in either of these is perceived as absence of hue.
Perception of a circle of hues is specific to organisms whose colour vision
involves three receptor (cone cell) types processed in a cone-opponent fashion,
that is, by comparing the responses of the three cone cell types with each other
at an early stage of visual processing. When the energy of an isolated light is
distributed throughout the spectrum in a balance similar to that of daylight, the
responses of our long-, medium- and short-wavelength (L, M and S) cone cell
types are mutually balanced and cancel each other out chromatically, and the
light is perceived as colourless , that is, "white" (Fig. 14.4.2A)1. Single
wavelengths on the other hand necessarily evoke an unequal response from the
cone cell types (Fig. 1.4.2B). Beginning at the short wavelength end these
unequal cone responses are successively S, S+M, M, M+L and L dominant. An
L+S dominant response, which completes the cycle of possible cone responses,
can be evoked only by lights containing mixed wavelengths predominantly from
the two ends of the spectrum.

Figure 1.4.3. (A) yellow/blue and (B) red/green colour-opponent hue

perceptions evoked by different wavelengths, based on the hue
cancellation experiments of Hurvich and Jameson (see Fig. 3.3, Section
3.2).

We perceive this cycle of possible cone responses as a cycle of hues comprising

successive combinations red, yellow, green and blue, apparently organized as a
yellow or blue and a red or green colour-opponent pairs (Fig. 1.4.3A,B). It
remains controversial how these colour-opponent perceptions (or "signals")
arise from the circuit of trichromat cone-opponent responses (Hardin, 2014),
but in general terms a yellow vs blue perception is evoked by long or
middle vs short wavelength dominance, while a red vs green perception is
evoked by long OR short vs middle wavelength dominance (Fig. 1.4.3B). Note
that by this model, "red" is not a property of light that our visual system detects,
it is a perception that our visual system creates in response to two distinct bands
of wavelengths that have nothing in common except that they occupy opposite
ends of the range visible to humans. For a dichromatic species having only two
cone types, say L and S cones, the range of relative responses is confined to just
L dominant and S dominant, which can be perceived as just two hues. Human
dichromats are believed to perceive colour something like this, though there is
evidence that colour perception in dichromatic-testing individuals of a
trichromatic species can be more complex.

Figure 1.4.4. Spectral hues explained as successive combinations of

yellow/blue and red/green colour-opponent hue perceptions.

The cone responses evoked by single wavelengths (S, S+M, M, M+L and L
dominant) are experienced as successive combinations of these colour-
opponent pairs, forming the the spectral hues spectral red, orange, yellow,
green, cyan, blue, and violet (Fig. 1.4.4A,B). Middle ("pure") yellow and blue are
experienced where the red/green perception is at zero, and middle green is
experienced where the yellow/blue perception is at zero. The non-spectral
S+L dominant response, requiring mixtures of wavelengths from the two ends
of the spectrum, is seen as the range of non-spectral hues from purple to
magenta and middle red. Broadband lights containing a mixture of wavelengths
evoke a cone response and perceived hue that depends on the overall balance of
wavelengths present.
Figure 1.4.5.Comparison of spectral reflectance curves of
Gamblin Cadmium Yellow Light, Cadmium Yellow Medium and Titanium
White oil paint. Munsell plot and spectral reflectance curves for Gamblin
Conservation Colours from drop2color by Zsolt Kovacs Vajna; paint
sample photos from dickblick.com.

The colours we see as belonging to objects at least to a degree relate to their

characteristic spectral reflectance, rather than to the wavelengths that they
happen to reflect under a particular illuminant, and the hues of these object
colours depend on the the direction of bias of these spectral reflectances. Paints
that strongly reflect the middle and long wavelength parts of the spectrum (Fig.
1.4.5) are seen as strongly yellow (Fig. 1.4.3A), and either slightly greenish,
slightly reddish, or neither depending on the relative size of the middle and long
wavelength contributions (Figure 1.4.3B). In Gamblin Cadmium Yellow Light
these contributions are approximately evenly balanced and the paint is seen as
middle yellow, neither greenish nor reddish, while in Gamblin Cadmium Yellow
Medium the contribution from the middle (green-evoking) part of the spectrum
is smaller, and so the paint colour is seen as having a red component (Fig. 1.4.5).
Titanium White is seen as being white because its spectral reflectance has no
substantial bias.

Despite a broadly based scientific consensus to the contrary, the view that hues
like red and green are properties or "things" physically residing in objects is still
alive and well in traditional colour theory accounts of "colour mixing", which
take the seemingly logical additional step of assuming that when we mix paints
the hues residing in the paints themselves mix. The traditional doctrine that red,
yellow and blue are primary colours that "can't be mixed from other colours"
proceeds from these assumptions: we can't make a red mixture without using
paint that already "contains red" (such as magenta), similarly we can't make a
yellow or blue mixture without using paints that already "contain" yellow and
blue respectively, but the colour green is not a primary colour but a mixture of
colours because we can make a green mixture from middle yellow and middle
blue paints that don't "contain" green. (For the scientific alternative to this
impeccable phenomenology see Section 5:Subtractive Mixing).

Unfortunately, many popularized explanations of colour vision also encourage

the idea that hues exist outside us by speaking of cone cell types that "detect red,
green and blue wavelengths" respectively (which is absolutely incorrect, See Fig.
1.4.2A), especially when they take this to mean that humans really only see red,
green and blue, that mixtures of red and green lights just "make you think
you're seeing yellow", or that the colour magenta/pink/purple alone is "made
up" by our brain, because it doesn't "exist" in the spectrum! Although the
colour-opponent model has been widely accepted alongside trichromacy in
vision science and psychology for many decades, most popularized explanations
of colour vision still omit it completely and explain only the trichromatic model
at the level of the retina.

Measuring Hue

Figure 1.4.6. The forty hue pages of the Munsell Book of Color (glossy edition).

Hue systems differ among themselves by the criterion used for placing hues
opposite each other, which affects the spacing of the hues, and by which hues
that are treated as "primary" or principal in anchoring the circular scale. More
incidental differences are the sequential direction of the hues (i.e. clockwise or
counterclockwise spectral sequence), whether the continuous loop of hues is
represented the form of a circle, hexagon, triangle, star, or other closed
geometrical figure, and the conventional orientation (i.e. a specific hue placed at
the top of the figure or treated as the first in the sequence). Hue systems are
classified on this site according to the criterion used to place hues opposite each
other, and hue systems based on the historical primaries and their
complementaries, opponent hues, additive complementaries, colourant-mixing
complementaries and perceptually-equal spacing respectively are discussed in
more detail on later pages.

Figure 1.4.7. A. Hue circle based on the historical primaries red, yellow
and blue from The Art of Color (1961) by Johannes Itten. The graphic
visually encapsulates Itten's view that his secondary colours green,
orange and violet each "contain" two of the primaries. B. Hue circle
based on opponent hue relationships from the Natural Colour System
(NCS) by the Scandinavian Colour Institute.

Hue systems based on the historical primaries are organized around the
conventional complementary pairs red-green, orange-blue and either yellow-
purple or yellow-violet. They include numerous 18th-21st century examples of
the "artists' color wheel" of traditional colour theory. These hue systems
reflect historical confusion of colourant-mixing and opponent hue relationships.

In traditional colour theory the concept of warm and cool colours is commonly
used to label distinctions of hue, for example warm yellow for reddish yellow
and cool yellow for greenish yellow. These warm/cool associations could be
considered a mild but widespread form of synesthesia. As is typical of that
condition, individuals who perceive warm/cool associations of colours can be
adamant that the association they perceive is objective and obvious, yet other
individuals can perceive the precisely opposite association. A striking example
concerns the colour blue, where different camps within traditional colour theory
regard a reddish blue like ultramarine as a warm and a cool blue respectively.
Even if one feels strong warm/cool associations of colours, in the interests of
clear communication it is wiser to speak of reddish or greenish blue rather than
warm or cool blue, and so on.
Hue systems based on opponent hue relationships are organized around
the opponent hue pairsred-green and yellow-blue, and include some historical
examples and the modern Natural Colour System or NCS (Fig. 1.4.7B). Different
individuals show a rather wide range of variation in the colour chips they select
as representative of the four opponent hues, but the averages for most studies
show a reasonable range of 2 or 3 Munsell pages (Fig. 1.4.8). These studies show
that the average determinations for the unique hues are not perceptually equally
spaced.
Figure 1.4.8. Munsell hue-chroma plane showing colours for sixteen
common artists' pigments as pure paints and as 1:1 mixtures with
Titanium white paint, calculated using Zsolt Kovacs-Vajna's
program drop2color. The coloured circles around the circumference show
average determinations of positions of unique red, yellow, green and blue
from four studies using Munsell chips cited by Kuehni (2012); studies
involving conversion from spectral data, NCS hues etc show further
spread.

Hue systems based on perceptually uniform spacing oppose hues that are
separated by an equal number of perceptual steps in either direction around the
hue circle. They include the Munsell system and CIE L*a*b* system. Broadly
speaking, opposite hue pairs in these systems are close to additive complements,
especially compared to other systems. The Munsell hue circle (Figs 1.4.6, 1.4.8)
is based on five principal hues, red, yellow, green, blue and purple (R, Y, G,
B and P) and five intermediate hues (YR, GY, BG, PB and RP). Each of these ten
hues was originally intended by Albert Munsell to have ten numbered divisions,
with the fifth division, for example 5R, being regarded as the typical version of
the hue. The Munsell Book of Color however has always had only four hue pages
for each principal and intermediate hue, for example 2.5R, 5R, 7.5R and 10R,
making a total of 40 hues.

Figure 1.4.9. Hue circle showing hue angle (H) used to specify hue in
HSB, HLS and HSI digital colour spaces. The circle arranges the additive
complementary pairs digital red-cyan (0o-180o) , yellow-blue (60o - 240o)
and green magenta (120o-300o) opposite each other. Note however that
opposite colours are not complementary away from these three axes. The
hue spacing is far from perceptually even, as can be determined both by
visual inspection and by the approximate positions of the Munsell
principal and intermediate hues (source).
Hue systems based on additive complements oppose hues of lights that make
white light when mixed. They include systems by Newton, Helmholtz, Rood and
Ostwald, and the CIE L*u*v* system. Examples of additive complementary pairs
are familiar to many from the major hue axes in the HSB hue system used in
graphics programs: the digital hues going by the names
of magenta/green, red/cyan, and yellow/blue. HSB hue angle (H) is calculated
from the ratios of the (nonlinear) RGB primaries by a simple formula, and is
highly uneven perceptually (Fig. 1.4.9).

Figure 1.4.10. Colour circle diagrams by Michel Jacobs from (A) The Art
of Colour (1923) and (B) The Study of Colour (1925).

Hue systems based on colourant-mixing complementaries oppose hues of

paints or dyes that make a neutral grey or black when mixed. The regular
succession of hues in the traditional artists colour wheel (e.g. Fig. 1.4.6A)
promotes the assumption that paint-mixing complementary pairs succeed each
other in an orderly succession of straight lines forming diameters of the hue
circle. Many painters evidently paint for years without questioning this
assumption, but in reality mixing paths show markedly different patterns for
paints that are close to the ideal subtractive primary hues compared to paints
that are far from these hues. Paints close to orange-red, yellow-green and blue-
violet in hue each neutralize or nearly neutralize paints of a remarkably large
spread of opposing hues, so that to make a hue circle that places paint-mixing
complementaries approximately opposite each other it is necessary to expand
the sectors occupied by these hues and contract the ranges of the remaining
hues. Michel Jacobs used a diagram of precisely this sort (Fig. 1.4.10) in his once
popular books The Art of Colour (1923) and The Study of Colour (1925), both of
which were republished in numerous editions until 1956.
Subdividing the Hue Page

Figure 1.4.11. Four approaches to dividing up a hue page.

In three-dimensional colour systems each hue corresponds to a two-

dimensional array of colours that can be called a hue page. Hue pages in
different systems, such as Munsell, NCS and digital hue angle H, typically drift
through each other to some extent. We've seen that hue pages in the Munsell
system are divided up according to value and chroma, but the same array of
colours can be classified in many alternative ways for different purposes (Fig.
1.4.11). Thus the digital colour space HSB divides up a hue page according to
measures of saturation and relative brightness (Fig. 1.4.11B), while the digital
colour space HLS divides up a hue page according to rather arbitrary measures
of relative lightness and relative chroma, confusingly called "lightness" and
"saturation" (Fig. 1.4.10D). Several important systems including the historical
Ostwald system and the modern Scandinavian NCS divide up a hue page (in
different ways) into white, black and coloured components (Fig. 1.4.11C).
1
These comments refer to light viewed in isolation; the perceived colour of a light or object can
of course vary considerably depending on surrounding, interspersed and recently
viewed colours.

Page added January 21, 2017.

<< 12 3 4 5 6 7 8 >>

1.5 The Dimensions of Colour: Chroma

  Chroma
 Measuring Chroma
 Measures of Relative Chroma

Chroma

Figure 1.5.1. Four plates displaying opposite hue pages from the
1915 Atlas of the Munsell Color System by Albert Munsell. A. 5BG and 5R
plate. B. 5B and 5YR plate. C. 5PB and 5Y plate. D. 5G and 5RP plate.

Chroma is the chromatic strength of an object colour, the perceived amount of

difference from a grey of the same lightness (value.) The term chroma (Greek
for "colour") was introduced by Albert Munsell as a dimension of his Munsell
system, in which it is represented by distance radially from the central value axis
along scales intended to be perceptually uniform across all hues (Fig. 1.5.1).
Various terms in different languages had of course been used for object colour
strength in older colour order systems, but all of these seem to have referred to
some form of relative chroma, that is, chroma relative to some amount deemed
to be the maximum for that hue, and typically represented as the radius of a
circular diagram.

The word saturation is often used loosely to mean chroma or some kind of
relative chroma, but is defined as an entirely distinct concept in CIE
terminology (Section 1.7).
Figure 1.5.2. Illustration of the CIE definition of chroma as the
colourfulness of an area judged as a proportion of the brightness of a
similarly illuminated white area. In the image at the top, the coloured
squares in each horizontal row (R1-R3 and P1-P3) have the same chroma
(and appear to be the same colour) because they each maintain the same
colourfulness relative to the brightness of the areas seen as being white
(W1-W3). (Note however that when R1-R3 and P1-P3 are judged against
the same white (such as the white surface of the graph), they are
perceived as different colours having different chromas).

The formal definition of chroma is based on the idea that when a chromatic
light-reflecting object is increasingly strongly illuminated, the colourfulness of
its appearance increases, but the brightness of a similarly illuminated white
object increases proportionately, so its intrinsic strength of colour or chroma
can be defined as the colourfulness judged relative to this brightness (Fig. 1.5.2).

Chroma: "colourfulness of an area judged as a proportion of the brightness of

a similarly illuminated area that appears white or highly transmitting"(CIE,
2011, 17-139).
Figure 1.5.3. Comparison of spectral reflectance curves of a medium- and
a high-chroma orange-yellow paint. Munsell plot and spectral reflectance
curves for Gamblin Conservation Colours from drop2color by Zsolt
Kovacs Vajna; paint sample photos from dickblick.com.

Chroma is how we perceive the absolute amount of spectral bias of an object's

reflectance/ transmission, that is, its efficiency as a spectrally
selective reflector/ transmitter of light. For an object to have high chroma it
must reflect/ transmit one or two parts of the spectrum very strongly and the
rest very weakly. In Fig. 1.5.3, Gamblin Cadmium Yellow Medium oil paint is
similar in hue to Gamblin Yellow Ochre, but has higher Munsell chroma (and
value) because it more efficiently reflects the wavelengths that contribute to an
orange-yellow appearance.

Figure 1.5.4. A. 5R hue page from the Munsell Book of Color, Glossy Edition. B. 5R
hue page for digital colours of a standard RGB colour space (sRGB) colour space,
and (shaded) limits of optimal colour stimuli (theoretical limits of colour for light-
reflecting objects).

If an object reflects all wavelengths about equally, whether at a very high level (a
white object) or at a very low level (a black object), it necessarily has very low or
zero chroma. If we could make a white object begin to absorb a part of the
spectrum, it would inevitably get darker as it got more strongly coloured;
similarly if we could make a black object begin to reflect part of the spectrum, it
would inevitably get lighter as it gets more strongly coloured. The potential
range of chroma thus increases as we move away from the black and white
extremes of lightness, up to a maximum at some intermediate value that
depends on the hue, being high for yellow and low for violet and blue (Fig. 1.5.5
below). The value at which maximum chroma is reached is sometimes known to
artists as the home value or the peak-chroma value (Gurney, 2010, p. 76). This
general pattern is repeated with variations in peak chroma and peak-chroma
value in the matte and glossy editions of the Munsell Book of Color, in digital
colours, and in the colour range of optimal colour stimuli, theoretical objects
that reflect 100% of one or two parts of the spectrum and 0% of the remainder,
and thus mark the theoretical limits of colour for non-luminous objects (Fig.
1.5.4).

Measuring Chroma

Figure 1.5.5. The forty hue pages of the Munsell Book of Color, Glossy
Edition. The peak-chroma value of these pages varies from 8 to 8.5 for the
5Y hue page to 3 to 4 for the 7.5PB hue page.
Figure 1.5.6. RGB colours of forty Munsell hues, after a diagram by Hans
Irtel (colours above chroma 18 not shown)

The scales of chroma in the Munsell system and in CIE L*C*h space (below) are
perceptually even, open-ended dimensions with no specific upper limit apart
from the theoretical limits of object colours. Munsell chroma is measured in
Munsell chroma units, judged by reference to these scales of physical chips in
the Atlas of the Munsell Color System (Munsell, 1915) and later the Munsell
Book of Color (1929-). Using the correlation with CIE colorimetry contained in
the 1943 renotation, Munsell chroma can be related to colorimetric
measurements such as CIE XYZ tristimulus values.
Figure 1.5.7. A, Chroma limits of each hue page of the Munsell Book of
Color, Glossy Edition (cf. Fig. 1.5.4). B, Chroma limits of mixing gamut of
a selection of Golden Heavy Body Acrylics, generated using the
program drop2color by Zsolt Kovacs-Vajna.

When Albert Munsell produced perceptually-even scales of colour chips for his
1915 Atlas he established that the maximum chroma he could obtain with his
paints not only occurred at different values for different hues, but also varied
considerably in absolute magnitude (Fig. 1.5.1). In the modern Munsell Book of
Color the highest chromas (16 Munsell chroma units) are attained in the hue
range from orange-yellow to orange-red, while the lowest maximum chromas
(10 Munsell chroma units) are reached in the hue range from cyan to green (Fig.
1.5.7A). A very similar pattern can be seen in the gamut of colours mixed from
any reasonably large range of artists' paints (Fig. 1.5.7B). This lopsided gamut of
colours available to painters, with a bulge between red and yellow, does not
really present a problem because the range of common object colours is
restricted in essentially the same way, for the same combination of physical and
physiological reasons.

Figure 1.5.8. Comparison of gamut of a range of artists' acrylic paints (in

colour) with that of standard sRGB digital colours (in grey). The paints
(Golden Heavy Body Acrylics) exceed the gamut of RGB colours in the
areas where the latter is relatively poor (around yellow and cyan), thanks
to the presence containing a good range of cadmium colours plus two
phthalocyanine greens (blue shade and yellow shade) and a
phthalocyanine blue.
The chroma limits of digital colours are very different however, far exceeding
artists' paints with Munsell chromas of 24 in the violet-blue to magenta range,
down to 18 at red (Figure 1.5.9A, below). Artists' paints on the other hand
exceed the gamut of standard (sRGB) digital colours where these are relatively
poor in the vicinity of yellow and cyan (Fig. 1.5.8).

Figure 1.5.9. Gamut of a standard RGB colour space (sRGB) in (A)

Munsell and (B) CIE L*a*b* space, in oblique perspective view (above)
and in plan view (below).

In several important colour spaces colours are arranged as in hue-lightness-

chroma spaces, but are specified using two chroma dimensions orthogonal (at
right angles) to each other, instead of hue and chroma. These orthogonal
dimensions, loosely related to the opponent axes red-green and yellow-blue, are
more amenable to calculating colour differences than the polar hue-chroma
dimensions. Position in the hue-chroma plane, whether specified by orthogonal
or polar dimensions, is sometimes called chrominance (Kerr, 2005). In CIE
L*a*b* colour space (Fig. 1.5.9B), widely used in colorimetry, and appearing
as Lab colour space in graphics programs such as Photoshop, colours are
specified by two chroma dimensions, +a/-a and +b/-b that coincide roughly
with the Munsell 10RP/10G and 5Y/5PB hue axes respectively. Thus instead of
being specified as having an orange hue at a certain chroma, a colour is specified
as having so much reddish chroma (a) and so much yellowish chroma (b). When
a more intuitive representation is required, CIE L*a*b* coordinates can be
converted very simply to measures of hue, lightness and chroma (L*C*h). The
measure of chroma (C*) is calculated by applying Pythagoras' theorem to the a
and b dimensions (square root of a2 + b2), and h is the hue angle measured
counterclockwise from the +a axis. A physical colour atlas of hue pages of
coloured chips is available that embodies this system (the RAL Design
Atlas). Other orthogonal colour spaces known as YCbCr, YUV and YIQ, are used
in television and video, and in jpg compression.

Figure 1.5.10. Adobe Photoshop colour picker, showing RGB colours with
Lab lightness (L) of 63. NB: many combinations of Lab values within the
square are outside the range of real RGB colours; available RGB colours
with L=63 are confined to the area here outlined in black.

In Photoshop colours can be selected using the orthogonal dimensions of Lab

space using either the sliders in the Color window or the Lab controls in the
colour picker. This permits colours to be chosen by lightness, but one trap is
that real RGB colours are confined to a vaguely visible zone towards the middle
of the square shown in Fig. 1.5.10. Clicking on the brightly coloured areas
outside this zone will yield colours that are the closest available, but which don't
actually have the Lab coordinates they are shown as having.

Measures of Relative Chroma

Dimensions of relative chroma judged either (1) as a proportion of the chroma

deemed to be the maximum possible for a given hue, or (2) as a proportion of
the maximum possible for a given hue at the same lightness, etc., were included
in several early colour order systems, including the early 19th century systems of
Klotz and Gregoire (see Kuehni and Schwartz, 2008). The concept of a
"maximum" chroma is potentially problematic as it depends on what
comparisons are ruled in or ruled out. If the maximum chroma is based on the
highest chroma attainable with dyes and pigments at a given time, this may
change with developments in chemistry. If it is based on estimates of colour
content by test subjects, it matters what the subjects' were thinking of as the
highest chroma colour. For example, did they include fluorescent-looking
colours or the even greater chroma of an area emitting bright monochromatic
light?

Figure 1.5.11. A. Denman Ross' colour classification from Snow and

Froehlich (1904) B. Three of the twelve hue pages, BV (blue-violet), V
(violet) and VR (violet red) from Denman Ross' The Painter's
Palette (1919). The vertical lines represent four degrees of "intensity",
while the oblique lines represent four degrees of "neutralization".

Albert Munsell's friend and rival Denman Ross presented a simple conceptual
classification of colours in terms of three attributes: hue ("color"), lightness
("value") and a measure of relative chroma he called intensity, judged relative
to the maxmum attainable for the same hue from a specified set of paints (Fig.
1.5.11). Ross also used a measure called neutralization for chroma relative to the
maximum possible for a given hue at a given value. Both intensity and
neutralization were to be expressed on a scale of 0, 1/4, 1/2, 3/4 and 1. Fig.
1.5.11B shows Ross' hue pages for three of the twelve hues in his system.
Figure 1.5.12. Horizontal (left) and vertical (right) cross sections of the
colour sphere described by Johannes Itten in his The Art of Color (1961)

In the color sphere described by Johannes Itten in his The Art of Color (1961)
each hue is represented by a tint-shade scale between white and black at the
surface and two intermediate degrees of relative neutralization between these
scales and the greyscale axis. (Fig. 1.5.12). In the simple digital colour
space HLS, so-called "saturation" is a similar dimension of relative
neutralization. The difference between these measures and Ross' neutralization
dimension is that the latter measures chroma relative to the maximum possible
at a given lightness, while the relative neutralization dimensions of Itten and
HLS measure chroma relative to the maximum possible a given hue and vertical
position in the system, which varies in lightness for different hues.

The Scandinavian NCS system has a dimension called chromaticness

representing chromatic content considered as a proportion of the colour as a
whole. It is related to both chroma and saturation as defined by the CIE, and is
discussed in a later section.

Page added January 23, 2017. A much earlier (2007) account of chroma is
located on this site here.

<< 12 3 4 5 6 7 8 >>

1.6 The Dimensions of Colour: Brightness and

Colourfulness
 Dimensions of Colours of Light
 Seeing Brightness
 Alternative Definitions and Usages
Dimensions of Colours of Light

Figure 1.6.1. Digital painting of a stripe of red paint passing between

shadow and light.

Consider a stripe of red paint passing between shadow and light (Fig. 1.6.1). By
virtue of the pre-conscious processes that equip our vision with a high degree of
object-colour constancy, we instantly, effortlessly and automatically see the
stripe as being the same colour, that is, as having the same red object colour,
over its whole length. This red colour can be specified more precisely in terms of
an object colour notation such as Munsell hue, value and chroma, and we could
confirm that the stripe matches the same Munsell chip placed beside it in the
shadow and in the light. Nevertheless, the appearance of the stripe
is brighter and more colourful in the light than in the shadow.

The CIE definitions of brightness and colourfulness are:

Brightness: "attribute of a visual perception according to which an area

appears to emit, or reflect, more or less light. NOTE The use of this term is not
restricted to primary light sources" (CIE, 2011, 17-111).

Colourfulness: "attribute of a visual perception according to which the

perceived colour of an area appears to be more or less chromatic" (CIE, 2011, 17-
233).

In our example of the red paint stripe the colour attributes of brightness and
colourfulness pertain to the perceived colour of the light reaching our eyes from
different parts of the stripe, rather than to the object colour seen as belonging to
the stripe itself. So while perceived colours of objects can be described in terms
of just three dimensions such as hue, lightness and chroma, perceived colours of
lights are described in terms of different sets of three dimensions, such as hue,
brightness and colourfulness, or alternatively as discussed in the next section,
hue, brightness and saturation. Describing the colour appearance of an
illuminated object involves both types of colour, and so requires more than
three dimensions.

Figure 1.6.2. A White's illusion: the grey bars all have the same physical
luminance, but are perceived to differ in brightness and lightness. B. Fig.
1.6.2A viewed through a screen with two apertures. C. Light-shedding
illusion by Akiyoshi Kitaoka from his website Akiyoshi's Illusion Pages:
the central area is identical in luminance to the white surrounding the
figure, but appears brighter and thus luminous. D. Fig. 1.6.2C viewed
through a screen with two apertures.

Brightness is our perception of luminance (visible light energy), whether of a

primary light source or of a light-reflecting or light-transmitting object.
Perception of absolute luminance is strongly influenced by the state of
brightness adaptation of the observer: for example, your laptop screen is
comfortably bright when viewed indoors but very dim outdoors, even though it
has exactly the same physical luminance. Perception of the relative luminance
of a given area in the visual field is influenced by the luminance of the
immediate surround (contrast phenomena) and of finely interspersed areas
(assimilation). Contrast phenomena (e.g. simultaneous contrast and Mach
bands) tend to push the brightness and other perceived colour attributes of an
area away from those of the immediate surround, while assimilation tends to
move the perceived colour attributes of an area towards those of finely
interspersed areas. In some carefully contrived examples like White's illusion
(Fig. 1.6.2A) and light-shedding illusions (Fig. 1.6.2C) effects of this kind can
result in particularly large misperceptions of relative luminance. Misperceptions
of this kind are not eliminated by "squinting" (which actually increases effects
involving assimilation like White's illusion), and in general observers can only
perceive relative luminance veridically (that is, as being equal in the areas
concerned) by using an opaque screen with two small apertures (Fig. 1.6.2.B,D).

Figure 1.6.3. Above: Sunrise, Blayney. David Briggs, 2014, oil on canvas,
12" by 24". Below: detail of Sunrise, Blayney viewed through a screen
with two apertures, showing approximate equality of brightness of the
areas depicting the sun and the brightest parts of the clouds.

In the landscape painting Sunrise, Blayney an attempt was made to use the
image colour relationships built into light-shedding illusions to create an
impression of luminosity in the area depicting the sun, even though this area is
painted with the same white paint as the brightest of the clouds on the left.

A variety of sophisticated quantitative measures of brightness and colourfulness

have been devised in modern colour appearance models, including the
Nayatani et al., Hunt and CIECAM02 models (see Fairchild, 2013 for detailed
accounts). Painters however measure variations in brightness and colourfulness
in a real or imagined scene in terms of the lightness and chroma of the paints
they will use to represent that scene. Indeed, looking a scene like that depicted
in Fig. 1.6.1, many painters would simply say that the lightness and the chroma
of the stripe increase from left to right, automatically thinking in terms of the
colours of the paints they would use. However the painter's representation of
brightness and colourfulness as lightness and chroma, like the photographer's,
involves choosing consciously or otherwise from a range of possible scaling
strategies, for example "crowding" the lights or the darks (Pope, 1922).

Seeing Brightness

Figure 1.6.4. Checker shadow illusion created by Edward H. Adelson.

(click to enlarge)
Source: http://web.mit.edu/persci/people/adelson/checkershadow_illusion
.html. Adelson's caption on the preceding web link ("The squares marked
A and B are the same shade of gray") slyly misdirects us.
The squares marked A and B "exist" only in the scene depicted in the
image, where they have different object colours; it is the rhomboidal
image areas labelled A and B that have the same colour and luminance,
difficult as it may be to attend to this until the representational spell of
the image is broken (right).

Perceiving brightness is not as straightforward a task as it might first appear. By

virtue of the pre-conscious processes responsible for object-colour constancy,
our perception of lightness is normally immediate, effortless and automatic. We
clearly also have an immediate sense of variations of brightness within the
visual field in a general way, but the task of making precise comparisons of
brightness between different parts of the visual field proves to be surprisingly
difficult. We discover this both when we are learning to paint and when we look
at optical illusions such as Adelson's well-know checker shadow illusion (Fig.
1.6.4, left). The image areas A and B have the same luminance, but because we
automatically see A as a dark surface and B as a light surface it is extraordinarily
difficult to see and compare the colours of the image in these areas. Even if we
consciously observe that B is a light coloured surface in a shadow and that A is a
dark coloured surface in light, and therefore rationally infer that the two areas
must at least be more similar in brightness than they are in lightness, it remains
stubbornly difficult to see this similarity, let alone to see that they are identical
in brightness. Our attention is so firmly held by these pre-consciously inferred
differences in lightness that we literally cannot see the identical brightness of
the areas until we break the spell of the image by connecting them up (Fig. 1.6.4,
right).

It is commonplace in teaching painting that beginning students paint objects in

something closer to their local colour than to the colour that they need to be to
look to have this local colour in the painting. As non-painters, our conscious
visual perceptions primarily track properties of objects like reflectance rather
than properties of the visual field like luminance. There is nothing peculiar
about this: in evolutionary terms the primary value of our visual system is in
providing information about objects in the world rather than raw information
about the visual field. Our visual system clearly collects information about
relative luminance accurately enough to permit pre-conscious "inference" of
reasonably stable object colour properties (however this is done!), but we are
not accustomed to accessing this relative luminance information consciously.
We do not need to do so until we start to paint.

Faced with this difficulty, painters have developed various strategies to help
them to make more accurate judgements of relative brightness. One of the best
known is "squinting", that is, viewing through the eyelashes of nearly closed
eyes. This has the effect of flattening the 3D model of object colours and
illumination that our brain has so cleverly disentangled for us into a blurred
two-dimensional array of light (watch how things jump back in space the instant
you open your eyes!). Alternative techniques that work for some observers
include staring at a fixed point and making comparisons using peripheral vision,
or letting the eyes wander casually around the visual field. ("unfocussing the
eyes"). With a little persistence at least one of these methods allows most
students to perceive (often quite suddenly) the equality of luminance of A and B
in Fig. 1.6.4. Even if all of these strategies fail, the use of a screen with two
apertures allows every student to perceive the relative luminance of the two
areas veridically.
Figure 1.6.5. Form illusion by R. Beau Lotto (click to enlarge).. The
two image shapes depicting tables have the same proportions measured
through the middle . For an interactive version of this and many other
illusions visit Lotto's website
at http://www.labofmisfits.com/articles/illusionsoflight.asp

An analogous situation exists with shape constancy. In Fig. 1.6.5 we can see
without difficulty that both of the rectangular tables depicted present
a trapezoidal shape in the visual field, but it is extraordinarily difficult to attend
to these areas as flat shapes in the picture plane enough to see that they have
exactly the same proportions measured through their centres. For
misperceptions like this induced by shape constancy it is the artists's strategy
of measuring with a pencil or paintbrush that comes to the rescue, allowing
proportions in the picture plane to be determined veridically.

Alternative Definitions and Usages

The current CIE definition of brightness explicitly states that the term is not
restricted to primary light sources. Previously the 1970 edition of the
CIE International Lighting Vocabulary (as quoted by Wyszecki and Stiles,
1982) had defined brightness as "the attribute of a visual sensation according
to which a given visual stimulus appears to be more or less intense, or
according to which the area in which a visual stimulus is presented appears to
emit more or less light". While the word "emit" here might be interpreted to
imply a primary light source, this clearly was not the intended meaning because
the same word is used in the accompanying definition of lightness as the extent
to which an area "appears to emit more or less light in proportion to that
emitted by a similarly illuminated area perceived as a 'white' stimulus".

Arend (1993) published more concise definitions of brightness and lightness

characterized by the author as "similar to the CIE and OSA definitions, but
slightly clarified and simplified". Brightness was defined as "apparent
luminance, the apparent amount of light coming from a visual direction" and
lightness as "apparent reflectance". The simplified definition again does not
restrict the term "brightness" to primary light sources.

In contrast, Marr (1982) and Palmer (1999) have used "brightness" and
"lightness"to refer to perceptions of primary light sources and of light-reflecting
objects respectively in the context of certain stages of visual perception. I am
grateful to Anthony Waichulis for drawing my attention to these non-standard
usages.

Figure 1.6.6. Test stimulus similar to one illustrated by Purves and Lotto
(2011), after Blakeslee and McCourt (2015) The four diamond-shaped
areas are identical in physical luminance and are perceived as similarly
light grey image colours (about 80 on the included scale), but due to
simultaneous contrast those on the darker background are seen by many
observers as slightly brighter, and lighter in image colour, than those on
the lighter background. The two diamond shapes on the darker
background in A and B appear a very similar or identical light
grey image colour, and thus are similar or identical in brightness in its
standard sense, but considered as object-colours depicted in the scenes
they differ greatly in lightness: light grey in A and much brighter than a
similarly illuminated white and thus self-luminous in B.

Purves and Lotto (2011, p.17) describe lightness as "the lighter or darker
appearance of object surfaces" and brightness as "the brighter or dimmer
appearance of objects that are sources of light (e.g. the sun, fire, lightbulbs
etc.)". From that point on they use the terms either interchangeably or
combined in the expression "lightness/brightness", and they discuss the
"peculiar relationship between lightness/ brightness and luminance" in a series
of examples, most of which are images depicting three-dimensional scenes like
Fig. 1.6.6. Lightness and brightness in its standard sense are simply correlated
only in two-dimensional patterns and in images of uniformly lit planes. In all
other images there is an important distinction between the proximal property of
image-colour lightness and the distal property of the perceived lightness of
objects depicted in the scene (Fig. 1.6.1, 1.6.6). What Purves and Lotto mean by
"lightness/ brightness" in images representing three-dimensional scenes is
therefore not obvious, but in their examples it seems to track an
overall impression of lightness or luminosity that is an unexamined conflation
of image-colour and object-colour lightness, aligning more closely with one or
the other according to how convincingly their image depicts a three-dimensional
scene. Unsurprisingly this impression of "lightness/ brightness" correlates very
poorly with physical luminance, and to account for its vagaries and those of
similar conflations of proximal and distal properties ("the angles between these
rods look different but they are actually the same" etc.), Purves and Lotto
propose a new empirical theory of vision in which visual perceptions
"correspond to the operational link between the stimulus and a successful
behavioral response, discovered over the course of evolutionary and individual
time". Their model assumes however that empirical input does not merely
influence our overall visual impressions but entirely defines all that we can ever
see.

As stated previously, the visual world is not a direct "window on reality" but is
more like a 3D computer model consisting of mind-generated perceived
attributes, including colour attributes of objects and illumination, projected
onto the external world. It's certainly reasonable to suppose that what
we ordinarily notice about this visual world is what has been useful to us in our
personal and evolutionary past, namely the unconsciously inferred properties of
objects. For this reason what we tend to notice in a convincing image of a three-
dimensional scene is the inferred lightness of objects in that scene and not the
lightness of the image itself. But there is an important difference between what
we ordinarily notice in the visual world and what we can learn to see.

Finally, there is a usage common both colloquially and in the textile industry in
which white, strong colours, and all intermediate tints are all described as
"bright". These are the colours that form the upward-facing surface of most
colour spaces including the Munsell system. Corresponding digital colours are
said to have a "brightness" B of 100 in HSB colour space because they each
represent the brightest possible digital colour for a given hue and saturation
(Fig. 1.7.6). In NCS terminology such colours have little or no blackness (Section
1.8) and in Evans' (1974) terminology they have approximately zero
greyness (Section 1.9).

Page added January 30, 2017.

<< 12 3 4 5 6 7 8 >>

1.7 The Dimensions of Colour: Saturation

 Saturation
 Measuring Saturation

Saturation
Figure 1.7.1. Three series of RGB colours each of uniform saturation: R1-R3 (high
saturation), P1-P3(medium saturation) and W1-W3 (zero saturation). Although
R1 and R3 have the same saturation, R3is seen as having higher chroma than
R1 when viewed in the same context, because by emitting a greater amount of
equally saturated red light, it is more colourful (but see also Fig. 1.5.2!).

The word saturation is often used loosely for chroma or some form of relative
chroma, but is defined by the CIE as a distinct attribute of perceived colour. In
Fig. 1.7.1 the series of RGB colours R1-R3 and P1-P3 each increase progressively
in brightness and colourfulness from left to right, and judged as object colours
they similarly increase in lightness and chroma. Yet each series is also perceived
to have a chromatic attribute in common: R1-R3 are perceived to emit
relatively pure red light, while P1-P3 are perceived to emit whitish red light in
which the white and red components are in a fixed balance. Within each series
colourfulness increases in step with brightness, but because of their greater
purity of light, R1-R3 are more colourful at a given brightness than P 1-P3.
Figure 1.7.2. Illustration of the CIE definition of saturation. R1-R3 have
high saturation, P1-P3 have moderate saturation and W1-W3 have zero
saturation.

Saturation as defined by the CIE is this attribute of the relative colourfulness of

an area independent of its brightness, which in effect is the relative freedom
from whitishness of the light it remits. Saturation in this sense is how we
perceive the spectral purity or amount of imbalance of the spectral power
distribution of a light, as far as this can be detected by our visual system with
just three types of cone cells.

Saturation: "colourfulness of an area judged in proportion to its

brightness" (CIE, 2011, 17-1136).

As saturation is defined as a function of two attributes of the colour appearance

of a single area, it depends entirely on the appearance of the light from that
area, and not on "judgements" in relation to other areas of the visual field, as do
the object colour attributes lightness and chroma. Since saturation and
colourfulness are not independent, we can describe colours of light either in
terms of hue, brightness and colourfulness or in terms of hue, brightness and
saturation.
Figure 1.7.3. Lines of uniform saturation on Munsell Book of Color,
Glossy Edition (A) 5Y and (B) 5PB hue pages. The 5PB chips attains
higher saturation but lower chroma than the 5Y chips.

Saturation differs from the other attributes of colours of light, brightness and
colourfulness, in that it remains essentially constant for a given object under
different levels of illumination unless the brightness is very high (CIE, 2011, 17-
1136, note). When a chromatic object is increasingly strongly lit by the same
illuminant, the spectral power distribution of the light it reflects should remain
the same while its brightness increases, and we would therefore expect the
saturation or perceived spectral purity of this light to tend to also remain the
same. Object colours can therefore be characterized by the saturation and
relative brightness of the light they reflect, as an alternatice to using their
chroma and lightness. Because chroma and lightness are colourfulness and
brightness judged relative to the same thing, their ratio reduces to the ratio of
colourfulness to brightness, that is, to saturation. So on a Munsell hue page,
colours of uniform saturation lie along lines that radiate from near the zero
point on the value (lightness) scale, in contrast to lines of uniform chroma
which are of course vertical (Fig. 1.7.3).
Figure 1.7.4. A. Seven uniform saturation series (including the
achromatic surround) perceived as uniformly coloured objects under
varying illumination. David Briggs, 2007, Photoshop CS2. B. Colours
from A plotted in YCbCr space. C. Colours from A plotted in L*a*b*
space. B and C plotted using the program Colorspace by Philippe
Colantoni.

Similarly, if the ratio of the RGB components of a series of digital colours is

maintained as their brightness increases, we would also expect their spectral
power distribution and saturation to stay the same (as each of the series in Fig.
1.7.1). Figure 1.7.4A shows how readily our visual system reads uniform
saturation series as an array of uniformly coloured objects under varying
illumination, provided that the arrangement permits such a reading. These
radiating lines of uniform saturation, in which chroma increases steadily in step
with lightness, are sometimes called shadow or shading series, and are very
important to painters because they define the paint or digital colours that create
the appearance of a uniformly coloured diffusely reflecting object turning under
a light source.

Measuring Saturation

Figure 1.7.5. Munsell hue pages of digital colours

from www.andrewwerth.com/color/. Quantified as chroma relative to
lightness, the maximum saturation attained by digital colours of different
hues varies greatly.

Different colour appearance models use a variety of formulae working from

different parameters to quantify saturation (Fairchild, 2013). The simplest
formulae quantify saturation as the ratio of chroma to lightness (RLAB, ZLAB).
Quantified in this way, different pages of the Munsell Book of Color attain
different maximum saturations, as well as different maximum chromas, on
different hue pages (Fig. 1.7.3, 1.5.5). Digital colours are even more varied in the
maximum absolute saturation they attain at different hues (Fig. 1.7.5, 1.5.6).
Figure 1.7.6. RGB components of digital colour P3 from Fig. 1.7.1. P3 has
an HSB saturation of 100 x 0.5 = 50.

The parameter called "saturation" (S) in HSB (or HSV) colour space, used for
example in the Adobe Photoshop colour picker, is a rough physical estimate of
the saturation of an RGB colour relative to the maximum saturation possible
for its Hue angle (H), and is based on a very simple calculation from its r, g and
b components (Fig. 1.7.6). For example, R100 G0 B000, R200 G0 B00 and
R100 G100 B000 are all fully saturated (S = 100), while R200 G100 B100 has a
saturation of 50. Remember however that the maximum saturation attained by
RGB colours varies greatly with hue (Fig. 1.7.5), and an RGB cyan of S=100
(G255 B255) is far less saturated (that is, more whitish) than an RGB blue of
S=100 (B255).
Figure 1.7.7. RGB colours of HSB Hue angle = 0 projected onto* the
Munsell 7.5R hue page, showing lines of uniform HSB saturation and
brightness. ( *Although the colours all have the same digital hue angle
(H=0), they drift somewhat in Munsell hue).

The parameter called "saturation" in HLS colour space is completely unrelated

to the standard definition of saturation, and is a kind of relative chroma
(chroma relative to the maximum possible for a given "L"). "Saturation" has yet
another two meanings in Photoshop in the definition of the Hue, Saturation,
Color and Luminosity layer modes. When these layer modes are used in Lab
mode "saturation" means Lab chroma , but when they are used in RGB mode it
is a measure of chroma in YCbCr space relative to the maximum possible for
RGB colours.
Figure 1.7.8. Is A the same colour as B or D? It all depends on how you
look at it, which is why we need to be so careful about terminology.
Viewed as objects in a scene, we see two strips, AB and CD, each of
uniform lightness and chroma, AB lighter and higher in chroma than CD.
Seen as image colours, however, A and B are surfaces of different
lightness and chroma, and emit light of different brightness and
colourfulness. They have been painted this way in order to represent the
brighter and more colourful light that a surface of uniform chroma
reflects where it is more strongly lit. Seen as light, all four areas emit light
of the same saturation - in each area, only the red phosphors are glowing
(right). Areas A and D emit light of the same hue,
saturation and brightness - that is, they are identical image colours, even
though they are perceived as very different object colours in the scene.
Indeed, because these perceived object colours are so visually insistent, it
may be very difficult to see these image colours as identical without in
some way breaking the representational spell of the image. Image: David
Briggs, 2007.

Page added January 30, 2017.

<< 12 3 4 5 6 7 8 >>

1.8 The Dimensions of Colour: Blackness and

Brilliance
 Black and White Content (Ostwald System)
 Blackness and Chromaticness (Natural Colour System)
 Brilliance
Black and White Content (Ostwald System)

Figure 1.8.1. Plate from one of Wilhelm Ostwald's colour atlases

(the Farbkoerper, 1919), with red arrows added to indicate increasing
black and white content. This blue-yellow plate is one of twelve plates
displaying 24 hues that together describe a colour space in the form of a
symmetrical double cone. In the Ostwald system the "Full" (most
chromatic) colour of each hue is located on the equator of the space,
whose vertical dimension therefore does not represent absolute lightness.

The CIE International Lighting Vocabulary does not claim that hue, brightness,
lightness, colourfulness, saturation and chroma are the exclusive or even the
most "natural" attributes of colour, but states only that perceived colour can be
described in terms of these six defined attributes. Another set of perceived
colour attributes, not defined by the CIE, consists of proportions of black, white
and colour content considered as components making up an object colour. This
way of perceiving object colours was discussed by Ewald Hering (1878
[translated 1964]), who illustrated it in the form of a veiling triangle with points
representing "completely pure white", "completely pure black" and the perfectly
"clear" hue (e.g. red). Object colours could be compared in terms of the
proportions of these components in various ways, for example in terms of
the relative degree to which the clear hue was veiled by a grey containing black
and white in a particular proportion. Hering noted several difficulties with using
these attributes to describe colour in absolute terms, especially that the clarity
of a chromatic colour cannot be determined exactly (ibid., p. 53).
Figure 1.8.2: A. Ostwald's "Analytical Triangle" in which the position of
a colour represents its implicitly physical content of black, white and full
colour, presumably as determined by spinning-disc mixture. Internal
lines show logarithmic subdivision of the white/black and the full
colour/white ratio that Ostwald believed would result in psychologically
equal spacing. B. Ostwald's "Logarithmic Triangle" showing the
divisions used in his published atlases (Fig. 1.8.1), in which the internal
divisions of the Analytical Triangle are spaced equally. The red and blue
lines each represent colours of equal saturation. Both diagrams from
Ostwald (1931).

The Ostwald system by Wilhelm Ostwald (1853-1952) was strongly influenced in

its geometry by Hering's model of colour perception, though its detailed
structure is based on spinning disc (additive-averaging) mixing. Ostwald
believed he had solved the problem of defining the perfectly clear or "Full
Colour" with his concept of a "semichrome", an optimal colour stimulus in
which 100% of all wavelengths between two complementary wavelengths are
reflected and all other wavelengths are completely absorbed. His "Analytical
Triangle" represents colours according to their (implicitly physical) content of
pure White (W), pure black (B) and "Full Colour" (C), but he considered that to
create psychologically equidistant colours it was necessary to scale the sides of
this triangle logarithmically in accordance with Fechner's Law. Equalizing the
size of these divisions produces the "Logarithmic Triangle" used in the plates of
his atlases, in which each colour has a two-letter designation representing
logarithmic black and white content, the third (chromatic) component being
redundant. In theory vertical series of colours in Ostwald's system
("isochromes") are uniform in saturation rather than chroma (Fig. 1.8.2).
Ostwald published his system in several colour atlases, including
the Farbkoerper of 1919 (Fig. 1.8.1), in which the Full Colour for each hue is
represented by the nearest colour to the semichrome attainable at the time with
colourants. Foss et al. (1944) showed that compromises in the ideal Ostwald
system are necessary to produce a physical atlas, and that Ostwald's actual
samples correspond poorly to the ideal Ostwald system. The influence of
Ostwald system in art education was high in the first half of the twentieth
century and continued to a lesser extent in the second half through the support
of Faber Birren among others.

Blackness and Chromaticness (Natural Colour System)

Figure 1.8.3. Screenshot of the online NCS Navigator, an interactive tool

for exploring the NCS system, and for selecting palettes based on
constant hue, whiteness, blackness, or chromaticness. This screenshot
shows on the left the overall double-cone geometry, which can be tilted at
any angle and filled in various ways, and on the right the one of the forty
hue pages that can be selected.

The Scandinavian Natural Colour System (NCS) like the Ostwald system
consists of triangular hue pages that together make up a double-cone space,
although unlike the Ostwald system this space is only partly filled with colour
samples. Colours are specified in terms of black content (blackness, s) and
chromatic content (chromaticness, c), both expressed as percentages, followed
by a hue designation. For example the designation 1050-R90B means the colour
with a blackness of 10 and a chromaticness of 50 on the R90B hue page. The
NCS differs from the Ostwald system in being based purely on colour
perception, and the scales of black, white and chromatic content are based
ultimately on averages of estimates reported by test subjects. The difficulty
subjects experience in establishing maximum chromaticness can be alleviated in
practice by showing them a highly chromatic sample of a specified c-value
(Seim, 2013). The NCS atlas has 40 hue pages of matte samples, in the current
edition ranging from 05 to 90 in blackness and up to between 55 and 85 in
chromaticness depending on the hue. Lightness is not a fundamental dimension
of the NCS system, but the current atlas incorporates lightness as variously
sloping contours on each hue page.

Figure 1.8.4. Lines of uniform blackness on the NCS Y50R, G50Y, B50G
and R50B hue pages, each projected onto the nearest Munsell hue page,
5YR, 5GY, 5BG and 7.5PB respectively (after Billmeyer and Bencuya,
1987). Colours represented in the Munsell Book of Colour Matte
Edition shown in colour (NCS chips are also matte); limits of optimal
colours shown as grey grid.

Billmeyer and Bencuya (1987) examined the relationship between the NCS
system and the Munsell system. Lines of uniform NCSÂ blackness descend
obliquely outwards on each Munsell hue page at an angle that varies according
to hue. The zero blackness line passes just above the highest matte Munsell
chips and thus describes an unevenly conical form in hue-value-chroma space.

Figure 1.8.5. Munsell value and chroma of uniform chromaticness series

on four NCS hue pages, Y50R, G50Y, B50G and R50B (after Billmeyer
and Bencuya, 1987).

Based on the analysis by Billmeyer and Bencuya (1987), lines of uniform NCS
chromaticness approximately track lines of uniform Munsell chroma (i.e. are
subvertical) for light colours, but swing around to approximate lines of uniform
saturation for colours of very low white content (Fig. 1.8.5).
Figure 1.8.6. A. Eight colour swatches of identical NCS blackness (30)
and chromaticness (30) from the NCS Digital Colour Atlas 1950 (2007). B.
Eight colour swatches of identical lightness and chroma, with hues
similar to, and Munsell value and chroma close to the average of, the
swatches in A. After an experiment by Green-Armytage (2006).

Because of the primary importance of lightness in image perception and image

making (Section 1.3), systems like the Ostwald and NCS that lack this dimension
have found much less favour with painters than lightness-based systems such as
Munsell and Lab space. The Ostwald and NCS systems however have had
considerable influence in various fields of design such as environmental design.
Sets of colours of equal NCS nuance (equal black/ white/ chromatic content) are
reported to be more aesthetically-pleasing for many observers than sets of equal
value and chroma (Fig. 1.8.6). In addition to the unity imparted by their shared
black/ white/ chromatic content, these sets of colours maintain the relative
lightness relationships shown by the full colours of each hue (yellow lightest,
etc.). Sets of colours related in this way have long been labelled a "concord", in
contrast to a "discord" where the "natural order" of lightnesses of the different
hues is disrupted (Carpenter, 1915).

Brilliance
Figure 1.8.7. Identical sets of three ellipses of varying luminance overlaid on three
differently illuminated areas of a photograph. In each case the dot that is close to
the brightness of the surrounding white paper looks like a bright or zero-blackness
yellow dot; the less luminant dots look olive-coloured, and the more luminant dots
look fluorent or luminous.

Hering and Ostwald stressed that dark colours like brown and olive are not
merely low in brightness, but have a distinct perceived attribute, blackness, that
is only experienced when bright colours are simultaneously seen in other parts
of the visual field. They also clearly understood that the perceived black content
of a stimulus is a function of its brightness relative to that of the surround, and
that as the relative brightness of a stimulus increases, the perception of
blackness diminishes and is replaced by the perception of luminosity. Ostwald
demonstrated this point by reference to an experiment by Hering that is well
worth doing personally:

"The only materials required are a piece of strongly colored paper (ideally
lemon yellow) and a larger piece of opaque white paper with an opening in the
center of about 2 cm diameter. Place the [yellow] paper horizontally on a
table, facing toward light and hold the white paper at a distance of 10 to 20 cm
above the yellow paper so that for the eye the colored one is completely
covered and the yellow color is only visible in the cut-out in the white paper. If
the white paper is kept horizontal, the resulting yellow appears normal, i.e.,
just like that obtained by direct viewing of the yellow paper. If its angle toward
the light is changed in a manner resulting in a steady increase of illumination,
whereby it is important to assure that no shadow falls on the yellow paper
below, the yellow in the opening changes. It becomes steadily grayer and its
color can, with a properly selected arrangement, that is, when the illumination
of the white paper in the optimal position is much stronger than that of the
yellow below, appear different all the way to a blackish olive-green. If the
white paper is angled in the other direction so that it is less illuminated the
yellow paper at first assumes its normal appearance and with increasing
shadowing of the white paper its appearance changes to a luminous light
yellow, a yellow that cannot be produced with pigments.

In this situation, the composition and intensity of the light rays reflected from
the yellow paper into the eye has not changed during the experiment; only the
surround in which the yellow color appears has changed in its brightness.
Nevertheless, the color of the objectively unchanged yellow has changed within
very broad limits and in particular has demonstrated all those changes that
can be experienced when the same yellow and a black are mixed in various
ratios on a disk mixture apparatus." (Ostwald, 1916, tr. Kuehni, 2010).
Figure 1.8.8. Illustration of Evans' experiments exploring the attribute of
brilliance. Observers saw in isolation a bipartite field consisting of a central
stimulus surrounded by a much larger achromatic stimulus that appeared white.
The luminance of the central stimulus could be varied without changing its
chromaticity. As its luminance was raised the central stimulus in turn reached the
threshold of blackness (A), a series of colours with perceived black content (B), and
then a colour that lacked perceived black content (C). With further increase in
relative luminance (suggested in this illustration by shading the field), perceptions
of fluorence (D) and eventually luminosity were reached.

In his book The Perception of Color (1974) and a series of earlier papers, Kodak
scientist Ralph Evans introduced the term brilliance for this scale from
blackness to luminosity. Evans devised a powerful variation on Hering's
demonstration, in which observers saw a bipartite field consisting of central
stimulus surrounded by a much larger achromatic stimulus that appeared white
(Fig. 1.8.8). The central stimulus could be adjusted through a very wide range of
luminances and chromaticities by using neutral and coloured filters, while the
luminance of the surround was held constant. Evans found that provided that
the central stimulus differs from the surround, this arrangement of just two
stimuli is sufficient to evoke not just brightness perception, but two distinct
additional perceptions, lightness (greyscale value) and brilliance. For a given
surround luminance, any luminance of the central stimulus below a certain
threshold (Evans' black point, B0) was perceived as being black, and increasing
the luminance above this threshold evoked decreasing perceived black content
(Evans' greyness) up to a point where perceived black content disappears
(Evans' zero greyness, G0). This point of zero black content is independent of
lightness, in the sense that it occurs at varying lightnesses for monochromatic
stimuli of different hues and for stimuli of the same hue and different
saturations. Increasing the luminance above this level results in the perception
of a fluorescent object colour (Evans' fluorence) and eventually the perception
of a light source. Evans (1974, p.167) considered that black content in the
Ostwald system was equivalent to brilliance, and that colours of zero black
content in the ideal Ostwald system were at his G 0.

Like lightness, brilliance is a perception of relative brightness, but whereas

lightness is brightness "judged" relative to the brightness of a similarly
illuminated white object, brilliance is brightness "judged" relative to the
maximum brightness possible or at least commonly encountered for a similarly
illuminated object of the same hue and saturation (Fig. 1.8.7). The word
"judged" should in neither case be taken to imply conscious reasoning:
perception of brilliance/blackness, like lightness, is immediate and automatic.
Brilliance is not currently defined or discussed in the International Lighting
Vocabulary, but the following provisional definition can be offered based on
Evans (1974).

Brilliance: relative brightness of an area judged on a scale of appearance

proceeding from black through decreasing blackness (or "greyness") to
fluorent (fluorescent-looking) and then self-luminous. Brilliance "may be
considered negative for grayness and positive for fluorence, or simply
continuous from the black point" (Evans, 1974, p. 100).

Lines of equal blackness descend steadily with increasing chroma for all hues to
a greater or lesser extent (Fig. 1.8.4). Consequently in each horizontal row on a
Munsell hue page brilliance increases to the right with chroma, least rapidly for
yellowish colours, most rapidly for bluish colours. This increase in brilliance is
perceived by many observers as an increase in brightness and lightness, an
effect known as the Helmholtz-Kolrausch effect. However students quickly learn
to distinguish this "colour glow" from luminance, and are able to correctly
match colours of equal luminance and Munsell value through the hue page
exercise in the New Munsell Student Colour Set and through practice matching
paint mixtures to a Munsell grey scale.
Figure 1.8.9. Munsell 5R digital hue page
from http://www.andrewwerth.com/color/ against white and black
backgrounds.

The perceived brilliance of an object, like its perceived lightness, is influenced

by the lightness of its immediate background. A Munsell N9 chip appears white
on a black background but appears light grey (that is, lower in lightness and
higher in blackness) when viewed on an N9.5 (lighter white) background. In a
Munsell hue page the highest chips in each chroma column appear "bright" with
little black content on a white surface, but the same chips appear distinctly
fluorent placed on a black surface (Fig. 1.8.9). In addition, most areas of the
visual field can be made to appear self-luminous if they are viewed in complete
isolation through a small aperture in a dark screen.

Figure 1.8.10. Beau Lotto's cube illusion. A video of this illusion can be
downloaded as "Cube I"
from http://www.labofmisfits.com/downloads.asp. Characteristically,
Lotto's video caption "See identical tiles differently" conflates image
properties with depicted object properties: the tiles "exist" only as
perceived objects in the scene depicted in the image; it is the rhomboidal
areas of the image that are physically identical. We find the blackish
brown and fluorent orange object colours seen as belonging to those
perceived tiles so visually insistent that we literally cannot see, that
is, direct our attention to, the similarity of the image colours until we
break the representational spell of the image by joining up the areas or
(as in the video) by masking the rest of the image.

What makes Beau Lotto's remarkable "Cube I" illusion (Fig. 1.8.10) so striking is
that the same psychophysical image colour is perceived as a black-containing
object colour and as a highly fluorent or luminous colour in different parts of
the image. The tile on the top face is much less bright than we would expect a
bright orange object to be based on the appearance of the other squares, so it is
seen as a chocolate brown object, but the tile on the shaded front face is brighter
than adjacent areas seen as white objects, so it is seen as a highly fluorent or
luminous orange object. Illusions like this show that of the two aspects of
physical reality, the two dimensional pattern of light making up the visual field,
and the three dimensional world of objects and illumination, our perception is
normally dominated by the latter, so it is the latter that dominates our attention
when we see convincing images of three dimensional scenes.

Page added February 14, 2017.

<< 12 3 4 5 6 7 8 >>

Part 2. BASICS OF LIGHT AND SHADE

SPECULAR AND DIFFUSE REFLECTION

Figure 2.1. Glossy sphere (billiard ball) under a single direct light source, showing effects of specular and diffuse reflection .
Photograph by David Briggs.

Light is reflected from most surfaces by two simultaneous processes,

known as specular (or surface) reflection and diffuse (or body)
reflection. Specular (literally "mirror-like") reflection creates
the highlight as its most conspicuous expression (Figure 2.1). On our
ball, as on most substances apart from the coloured metals like copper
and gold, the specular reflection retains the colour of the (white) light
source. The diffuse reflection is seen as the local colour of the surface
(in this case, off-white).

In specular reflection, light bounces according to the rule that the angle
of incidence (measured against a line perpendicular to the surface)
equals the angle of reflection. On any shiny object we see the highlight,
or specular reflection of the light source, at the point where the surface
is at just the angle needed to bounce light from the light source to our
eye in accordance with this rule (Figure 2.2). Consequently the
apparent position of the reflection varies with the location of the
observer, and even between the two eyes of the observer. On polished
objects our stereoscopic vision traces the two rays to our eyes back to
an illusory or virtual image that appears to be located, as in a
mirror, below the reflecting surface. Polished objects may show not
only the conspicuous image of the main light source, but also
recognizable reflections of the entire environment. Less polished
surfaces show a relatively "fuzzy" highlight, as the specular reflection is
spread over an area in which some microfacets lie at the critical angle.

Position of highlight. Light bounces to the viewer's eye at the point on

Figure 2.2.
the surface where the angle of incidence equals the angle of reflection, and the
viewer sees the highlight along this line. The viewer can not possibly see the
highlight at the point facing the light source; the light ray hitting the surface at
this point bounces straight back to the light source.

Many art instruction texts erroneously show the position of the

highlight at the point directly facing the light source (Figure 2.3). In
fact, the light ray incident at this point hits the surface at a right angle,
and bounces straight back to the light source, not to the observer. Only
in the case of a light source in the same direction as the observer does
the highlight fall at the point facing the light source. On a sphere the
highlight is always seen somewhere on the line between the point
facing the light source and the visual centre of the sphere.

Figure 2.3. Diagrams of light zones on a sphere from two popular art instruction texts.
Left from Loomis
(1951), and right from Aristides (2007). Both works show the highlight
(incorrectly) at the centre of the full light, facing the light source. As seems
conventional in this sort of diagram, the diffuse reflection of light from the
tabletop is shown but not the specular reflection, and a continuous band of core
shadow is shown alongside the terminator. The actual appearance of a sphere in
this lighting may be quite different (e.g. Figure 2.1).

The diffuse reflection consists of light that does not (macroscopically)

obey the rule of angle of incidence equals angle of reflection, but
instead is reflected equally in all directions. Several "popular science"
internet sites currently perpetuate the error that specular and diffuse
reflection result from surfaces being polished or rough respectively
(e.g. Molecular Expressions). Actually, diffuse and specular reflection
operate simultaneously from most materials, and evidently by quite
distinct processes, in that the specular reflection typically retains the
colour of the light source, while the diffuse reflection is often coloured
by the material. Diffuse reflection is considered to result from
subsurface scattering; that is, light enters a surface layer, is scattered
multiple times, and then exits in a random direction (Hanrahan and
Kruger, 1993). The exact proportion of the two kinds of reflection
depends essentially on the type of material (metals exhibit high
specular and no diffuse reflection), and on the angle of incidence (light
striking at a low angle to a surface is more likely to be reflected
specularly), but not on the roughness of the surface. Polishing a rough
surface will make a "fuzzy" specular reflection more concentrated by
reducing variation in microfacet orientation, but will not necessarily
increase the total amount of specular reflection unless it involves
removal of a diffuse-reflecting coating.
Subsurface scattering in some materials is extensive enough to cause
visible subsurface light transport, resulting in macroscopic
translucency. In some current literature on rendering in 3D
programmes, the term "subsurface scattering" is restricted to such
materials (milk, marble, etc.). Many other materials permit enough
subsurface scattering to create a diffuse reflection, but look opaque at a
macroscopic level, and only reveal their translucency in thin section.
Materials that are opaque even on a microscopic scale have no diffuse
reflection, and appear either black, sub metallic (e.g. graphite) or
metallic, depending on the amount of specular reflection they exhibit.

The amount of specular reflection generally appears to be greater on

dark objects than on white objects of similar composition, but this is an
illusion caused by swamping of the specular reflection by the greater
diffuse reflection from the latter.

Away from the highlight, specular reflection of light from the

environment occurs over the entire surface. This addition of specularly
reflected light, retaining as it does the colour of the environment, has
the effect of desaturating the colour of light reflected by coloured
objects. In general these specular reflections are most conspicuous in
the shadow zone, where they are not swamped by diffuse reflection,
and on the receding planes, where the specular reflection, being
created by light from behind the object hitting the surface at a low
angle, is more intense. The tendency to relatively strong specular
reflection on the receding planes may be quite noticeable even in fully
lit areas however, especially if there are light-toned objects in the
background.

<< 1 2 3 >>

Part 2. BASICS OF LIGHT AND SHADE

SPECULAR AND DIFFUSE REFLECTION

Figure 2.1. Glossy sphere (billiard ball) under a single direct light source, showing effects of specular and diffuse reflection .
Photograph by David Briggs.

Light is reflected from most surfaces by two simultaneous processes,

known as specular (or surface) reflection and diffuse (or body)
reflection. Specular (literally "mirror-like") reflection creates
the highlight as its most conspicuous expression (Figure 2.1). On our
ball, as on most substances apart from the coloured metals like copper
and gold, the specular reflection retains the colour of the (white) light
source. The diffuse reflection is seen as the local colour of the surface
(in this case, off-white).

In specular reflection, light bounces according to the rule that the angle
of incidence (measured against a line perpendicular to the surface)
equals the angle of reflection. On any shiny object we see the highlight,
or specular reflection of the light source, at the point where the surface
is at just the angle needed to bounce light from the light source to our
eye in accordance with this rule (Figure 2.2). Consequently the
apparent position of the reflection varies with the location of the
observer, and even between the two eyes of the observer. On polished
objects our stereoscopic vision traces the two rays to our eyes back to
an illusory or virtual image that appears to be located, as in a
mirror, below the reflecting surface. Polished objects may show not
only the conspicuous image of the main light source, but also
recognizable reflections of the entire environment. Less polished
surfaces show a relatively "fuzzy" highlight, as the specular reflection is
spread over an area in which some microfacets lie at the critical angle.

Position of highlight. Light bounces to the viewer's eye at the point on

Figure 2.2.
the surface where the angle of incidence equals the angle of reflection, and the
viewer sees the highlight along this line. The viewer can not possibly see the
highlight at the point facing the light source; the light ray hitting the surface at
this point bounces straight back to the light source.

Many art instruction texts erroneously show the position of the

highlight at the point directly facing the light source (Figure 2.3). In
fact, the light ray incident at this point hits the surface at a right angle,
and bounces straight back to the light source, not to the observer. Only
in the case of a light source in the same direction as the observer does
the highlight fall at the point facing the light source. On a sphere the
highlight is always seen somewhere on the line between the point
facing the light source and the visual centre of the sphere.

Figure 2.3. Diagrams of light zones on a sphere from two popular art instruction texts.
Left from Loomis
(1951), and right from Aristides (2007). Both works show the highlight
(incorrectly) at the centre of the full light, facing the light source. As seems
conventional in this sort of diagram, the diffuse reflection of light from the
tabletop is shown but not the specular reflection, and a continuous band of core
shadow is shown alongside the terminator. The actual appearance of a sphere in
this lighting may be quite different (e.g. Figure 2.1).

The diffuse reflection consists of light that does not (macroscopically)

obey the rule of angle of incidence equals angle of reflection, but
instead is reflected equally in all directions. Several "popular science"
internet sites currently perpetuate the error that specular and diffuse
reflection result from surfaces being polished or rough respectively
(e.g. Molecular Expressions). Actually, diffuse and specular reflection
operate simultaneously from most materials, and evidently by quite
distinct processes, in that the specular reflection typically retains the
colour of the light source, while the diffuse reflection is often coloured
by the material. Diffuse reflection is considered to result from
subsurface scattering; that is, light enters a surface layer, is scattered
multiple times, and then exits in a random direction (Hanrahan and
Kruger, 1993). The exact proportion of the two kinds of reflection
depends essentially on the type of material (metals exhibit high
specular and no diffuse reflection), and on the angle of incidence (light
striking at a low angle to a surface is more likely to be reflected
specularly), but not on the roughness of the surface. Polishing a rough
surface will make a "fuzzy" specular reflection more concentrated by
reducing variation in microfacet orientation, but will not necessarily
increase the total amount of specular reflection unless it involves
removal of a diffuse-reflecting coating.
Subsurface scattering in some materials is extensive enough to cause
visible subsurface light transport, resulting in macroscopic
translucency. In some current literature on rendering in 3D
programmes, the term "subsurface scattering" is restricted to such
materials (milk, marble, etc.). Many other materials permit enough
subsurface scattering to create a diffuse reflection, but look opaque at a
macroscopic level, and only reveal their translucency in thin section.
Materials that are opaque even on a microscopic scale have no diffuse
reflection, and appear either black, sub metallic (e.g. graphite) or
metallic, depending on the amount of specular reflection they exhibit.

The amount of specular reflection generally appears to be greater on

dark objects than on white objects of similar composition, but this is an
illusion caused by swamping of the specular reflection by the greater
diffuse reflection from the latter.

Away from the highlight, specular reflection of light from the

environment occurs over the entire surface. This addition of specularly
reflected light, retaining as it does the colour of the environment, has
the effect of desaturating the colour of light reflected by coloured
objects. In general these specular reflections are most conspicuous in
the shadow zone, where they are not swamped by diffuse reflection,
and on the receding planes, where the specular reflection, being
created by light from behind the object hitting the surface at a low
angle, is more intense. The tendency to relatively strong specular
reflection on the receding planes may be quite noticeable even in fully
lit areas however, especially if there are light-toned objects in the
background.

<< 1 2 3 >>

THE ZONE OF LIGHT

Figure 2.4. Glossy sphere under a single direct light source, showing
terminology of light and shade. Photograph by David Briggs.

Light from a light source travels in straight lines and divides any
subject into a zone of light and a zone of shadow (Figure 2.4). The zone
of shadow is further divided into a form shadow - an area in shadow
because it is turned away from the light source, and a cast shadow - an
area in shadow because the light source is blocked by another object.

The boundary between the zone of light and the form shadow is known
as theterminator. On spheres the zone of light occupies roughly half
the sphere (less than half if the light source is small and close; more
than half if the light source is large in apparent or angular size). In any
case the terminator on a sphere considered as a whole is a circle, and so
the visible part generally has the apparent shape of a section of an
ellipse. (In Figure 2.4 the light comes from almost exactly side-on to
the sphere, and so the terminator appears as an almost straight line).
The pattern of light and shadow visible depends on the relationship of
the direction of the light source to the direction of the viewer.
Experiment with the sliders in Figure 2.5 to see the effect of the
position of the light source on the shape of the light and shadow zones.
The character and direction of the terminator depends strongly on the
form of the object. Because of the steady curvature of the surface, the
terminator appears as a relatively soft edge on a sphere, an ovoid, a
cylinder or a cone (though becoming progressively crisper towards the
apex of the latter). The direction of the terminator runs on a sphere at
right angles to the direction of light fall, on a cylinder parallel to the
central axis, and on a cone radiating from the apex. On a prismatic
object the terminator tends to follow plane breaks, and is crisp in
proportion to the sharpness of the latter. On objects that are
intermediate in form, the terminator follows a path that is a
compromise between these basic patterns.

Figure 2.5 Effect of direction of light source. Drag the slider around the sphere
to see the effect of changing the direction of a close light source on the shape of
the light zones and the position and strength of the highlight, generated by Ray
Kristanto using Maya. Copyright David Briggs and Ray Kristanto, 2007.

Variation in the brightness of the diffuse reflection within the zone of

light creates the modelling of form. Irrespective of the position of the
observer, the diffuse reflection is always brightest on the plane directly
facing the light source, where light strikes the surface at 90 degrees.
Outward from this point, light hits the surface at progressively greater
inclinations, so that there is less incident light per unit area, hence less
light reflected, and hence a darker appearance. However this fall-off is
at first almost negligible, and the surface must curve away fully 60
degrees before the amount of incident light energy is halved; beyond
this point the fall-off of light is increasingly rapid. Artists find it useful
to divide this continuum into a series of light zones, comprising at least
a full light , a halflight, and commonly a dark halflight alongside the
terminator, and a centre light for the very lightest part of the full light.
When a painter has found the right colours for each of these zones, plus
the shadow, finding the transitional colours is easy, and even optional.

<< 1 2 3 >>

THE ZONE OF SHADOW

The shadow regions of a subject are often full of subtle variations, but
virtually all beginning students make too much of these. It is all too
easy to forget that we are looking at a shadow as soon as we begin to
observe the intriguing details within them. As always, it is a matter of
seeing these variations in the context of the total tone and colour range
of the subject.

Sources of illumination in the shadow zone of an object include light

reflected from the environment and any secondary light sources.
Reflected light from the environment is in turn generally visibly
reflected by an object in both a diffuse manner (the
conventional reflected light) and a specular manner. Although teaching
diagrams conventionally ignore the latter (e.g. see Figure 2.3), all
surfaces that are shiny enough to show a strong highlight will be shiny
enough to show visible specular reflections of the environment in the
shadow zone. The dark shape of the reflected cast shadow in particular
is often quite conspicuous, as in Figure 2.4. The diffuse reflection has a
variable pattern depending on where the form shadow surface is close
to, and faces, a strongly lit part of the environment. In some situations
this diffuse reflection may be brightest opposite the direction of the
main light source, creating a dark zone, often called a core shadow,
adjacent to the entire length of the terminator. In Figure 2.4 on the
other hand, the strongly lit tabletop creates a quite different pattern.

The cast shadow of an obliquely illuminated sphere is elliptical in plan

view, and remains persistently close to a perfect ellipse in appearance
when viewed in almost any perspective. The outer boundary of the cast
shadow is a transitional zone called the penumbra, formed where the
light source is partly obscured by the sphere. The width of the
penumbra, and hence the softness of the shadow edge, depends on the
distance between the light rays from either side of the light source (Fig
2.6); it increases with distance from the object casting the shadow, and
with the angular or apparent size of the light source.

Width of penumbra. The width of the penumbra generally increases with

Figure 2.6.
increasing distance from the object casting the shadow.

In most circumstances the darkest part of the shadow zone is

the crevice shadow, where two surfaces are in contact (e.g. the sphere
and tabletop in Figure 2.4). The darkness of this zone is the result
of occlusion - the further we go into the crevice, the less of the
environment can contribute light. Both the sphere and the tabletop are
darkened in the crevice shadow, although, depending on the direction
of observation, the darkened surfaces may not both be visible.

Experiment with the sliders in Figure 2.7 to observe the effect of

different combinations of two small and one large or ambient light
source. Multiple primary light sources will each create their own
pattern of light and shade, and these patterns combine additively in
terms of the energy of light coming from each point. Perceived
brightness however does not have a linear relationship to light energy.
Two equal light sources do not look twice as bright as one, and so if
overlaps of shadows get no light, they end up being visually very
conspicuous compared to the partly lit areas. Beginning students are
always very impressed by these odd-shaped dark overlaps, and copy
them faithfully, often making them even darker. In a painting or a very
extended drawing an artist may have time to capture the subtle effects
of multiple light sources, but otherwise it is usually advisable to
simplify by mentally filtering out the other sources, and drawing in
relation to one light source.

Figure 2.7. Multiple light sources.

Sliders vary the intensity of one large white overhead light
source (topmost slider) and two small yellowish oblique light sources. Copyright
David Briggs and Ray Kristanto, 2007.

<< 1 2 3 >>

PART 3. SOME BASICS OF COLOUR VISION

All art students are aware of the concepts of the colour wheel, the
spectrum, and the primary colours. They may even know just enough
to get into debates over whether the "real" primary colours are the red,
yellow and blue of the conventional artists' colour wheel, or the yellow,
magenta and cyan printer's primaries, or perhaps the three optimal
primaries for mixing coloured lights - red, green and blue. But few have
reflected on the apparent inconsistencies that lurk within this widely
held knowledge. For example, few have wondered why there are three
primary colours. If the visible spectrum consists of a continuous range
of wavelengths, why should just three colours be special? Why not five?
And why for that matter do colours form a circle, when the spectrum
does not? Though fundamental to the subject, these questions do not
normally occur to us, perhaps because we internalize our knowledge
about colour at an early and uncritical age. To answer them we need to
understand some basics of colour vision.

In broad outline, colour vision is currently understood to involve three

major processes:

1. Trichromatic input: information is recorded by the responses of

the L,M and S cone cells in the retina.

2. Opponent output: responses from the L,M and S cones are

converted into signals for yellowness vs blueness and
redness vs greenness, plus total brightness.

3. Processing for colour constancy: colour information from

throughout the visual field is analysed and resolved into an
interpretation of object properties (seen as their hue, value and
chroma) and lighting properties (seen as the hue, brightness and
saturation of the illumination).

We will review these processes briefly in the succeeding pages, not only
to answer the questions raised in the first paragraph, but also to gain
the background we need in order to discuss the dimensions of colour in
detail. For a much more comprehensive online account of current
understanding of all aspects of colour vision, the Webvision site of the
University of Utah is strongly recommended. For a short video
outlining these processes in the simplest possible terms, please see my
entry in the 2014 Flame Challenge (Section 3.6).

Unsurprisingly these processes are entirely ignored in "traditional"

colour theory, but unfortunately they are also quite misleadingly
treated in the majority of popular scientific explanations of colour.
Such explanations very generally follow a standard pattern that
includes a version of the trichromatic theory of input from three cone
types, but completely omits the opponent theory of output of red/green
and yellow/blue colour signals. Instead, these explanations talk as if
colours such as red and yellow exist within the spectrum itself , and
often as if individual cone types "detect" individual colours existing in
the spectrum (namely red, green and blue). These misunderstandings
have in turn generated various secondary misunderstandings about
colour, for example that since "real" yellow exists in the wavelengths of
the spectrum that look yellow, the yellow of a computer screen, which
contains few or no such wavelengths, is merely the "illusion" of yellow,
or that magenta alone is not a "real" colour, because it does not "exist"
in the spectrum. Two YouTube videos that have had the effect of
propagating these and other misunderstandings about colour, Michael
Stevens' This is Not Yellow and Steve Mould's The Mystery of
Magenta, are examined in detail in Section 3.7.

<< 1 2 3 4 5 6 7>>

TRICHROMACY AND OPPONENCY

Figure 3.1. Normalized absorbance of the three human cone pigments,
indicating the relative sensitivity of the three cone types to light of different
wavelengths. Image source: Maxim Razin, after Bowmaker J.K. and Dartnall
H.J.A., "Visual pigments of rods and cones in a human retina." J. Physiol. 298:
pp501-511 (1980). http://commons.wikimedia.org/wiki/Image:Cone-
response.png

Why three primary colours? The number three comes ultimately from
the fact that our colour vision begins with three types of light-sensitive
receptors, called cone cells, in the retina of the eye. The three cone
types have broadly overlapping ranges of sensitivity, and are
designated L, M and S (long, middle and short) according to the
relative position of their peak sensitivities in the visible spectrum
(Figure 3.1).

The idea of three receptor types was first proposed by Mikhail

Lomonosov and George Palmer in the eighteenth century in connection
with the then widely held belief that white light is a physically
trichromatic mixture of red, yellow and blue rays. In its modern form
the trichromatic theory stems from Thomas Young who, accepting the
innumerable spectral gradations of Newton, argued that these must be
detected by a finite number of receptor types, and suggested the
number three corresponding initially to the three artist's primaries,
and later to the red, green and violet spectral primaries from which all
other spectral colours could be obtained by additive mixing. In the
mid-nineteenth century Hermann von Helmholtz finally demolished
the theory of physical trichromacy and, together with James Clerk
Maxwell, revived and greatly developed Young's suggestion.
The L, M and S cones are sometimes described as red-, green-, and
blue-sensitive respectively, but this is misleading on several levels.
While the "blue-" and "green-sensitive" cones have their peak
sensitivities in the violet-blue and green parts of the spectrum
respectively, the so-called "red-sensitive" cones have their peak
sensitivity in greenish-yellow, not red. More fundamentally, individual
cone cell types do not "detect" colour, or even wavelength. The L and M
cones are sensitive to all wavelengths of visible light, and while each
responds to different parts of the spectrum to different degrees, their
response does not in itself encode which part of the spectrum the
stimulus came from. For example, an L cone responds in exactly the
same way to light of wavelengths of 630 nm (red), 475 nm (cyan) and
400 nm (violet) (Figure 3.1). Only the balance of L, M and S responses
to a monochromatic light serves to establish its position in the
spectrum. The colour of a such a light is a mental perception created by
our visual system based on post-receptoral comparison of the cone
responses. For example, redness (in an isolated light) is a perception
evoked by the predominance of the L and/or S cone response over the
M response. Once we consider a field of view with several lights, or
object colours, the relationship of L, M and S cone responses to
perceived colour is no longer direct.

Thus while it might seem common sense to believe that redness,

greenness and blueness are physical properties residing in different
parts of the spectrum, and that cone cells "detect" these properties, this
view is not compatible with the trichromatic theory of colour vision.
Colour is a psychophysical property of lights and objects that exists
only in relation to the visual system (just as sound is a psycholophysical
property correlated with the physical property of frequency of vibration
of air). For mixed lights, the same perceived colour can be evoked
many physically different mixtures of wavelengths (called metamers)
that have nothing in common except that they evoke the same balance
of L, M, and S responses in the human visual system..

What we can say is that the three cone types effectively divide the
visible spectrum into three bands - red-orange, green and blue-violet -
in each of which the response of one cone type predominates over the
other two (Figure 3.1). This three-fold division is visually evident in the
spectrum (Figure 3.2), and was recorded in the rainbow in antiquity by
Aristotle in his Meteorologica. The practical importance of these bands
is that if we take three light sources, one from each band, we can make
light mixtures with every possible combination of strong L, strong M
and/or strong S signals, and thus make strongly coloured mixtures
through the complete 360o range of possible hues. Three such lights
thus make effective primaries for additive colour mixing.
Figure 3.2. Solar spectrum. The orange-red, green and violet-blue bands, in
which the responses of the L, M and S cones respectively predominate over the
other two, are clearly evident on visual inspection of spectra such as this
complete solar spectrum. Image source:
http://www.adlerplanetarium.org/cyberspace/sun/learning.html (Credit: Nigel
Sharp, NOAO/NSO/Kitt Peak FTS/ AURA/NSF).

While the three-receptor theory successfully explained additive colour

mixing, it struck a problem in explaining our experience of colour.
Young and Helmholtz assumed that the three receptors were
responsible for "fundamental sensations" of red, green and violet, and
that other colour sensations such as white and yellow were "mixed"
sensations compounded from these. But as Ernst Mach and Ewald
Hering soon pointed out, there are indeed unique hues in our
psychological experience from which all others are compounded, but
these hues are red, yellow, green and blue, not red, green and violet.
These four hues are the only ones that we can experience as unique,
that is, unmixed with other colours - all others seem to be mixtures of
two unique hues. For example, we can experience or imagine a red that
is neither orangeish or purplish, whereas all orange colours are both
yellowish and reddish. In addition, these four colours are arrayed in
pairs, such that no colour can be red and green, or yellow and blue, at
the same time. Hering therefore rejected Helmholtz’s trichromatic
theory, and proposed instead his opponent theory, in which receptors
in the retina generate three opponent signals of red vs green,
yellow vs blue and white vs black.
However the opponent theory was not so successful in explaining
additive mixing: mixing unique red and green lights does not make
white light, as Hering had assumed, but yellowish light; to make white
light, either the red or the green must be somewhat bluish. Only yellow
and blue are additive complementaries as well as opponent hues. It was
realized early on that both Helmholtz and Hering could be essentially
correct if they were describing different stages of the colour vision
process, but most textbooks treated Hering's theory as a secondary
alternative until the 1950's when Leo Hurvich and Dorothea Jameson
devised a way of quantifying the red/green and yellow/blue
components of the colours of the spectrum (Figure 3.3). Hurvich and
Jameson proposed a two-stage zone model of colour vision, in which
responses from Helmholtz’s three receptors are compared with each
other to produce Hering’s three opponent signals. When studies of
goldfish and primates soon discovered opponent cells whose signals
were excited and inhibited by the responses of different cone types, it
seemed that the mechanism for Hering’s opponent signals had been
found, but these cone-opponent signals (L vs M and S vs (L+M)) turned
out not to coincide with the colour-opponenthue signals of colour
perception. Apparently therefore, the information encoded in the cone-
opponent signals is reshaped at a (poorly understood) later stage into
the red vsgreen and blue vs yellow colour-opponent signals. (This
reshaping could well explain why additive complementary and
opponent relationships differ for some hues).Multistage zone
models like this had already been proposed in 1930-40, but had been
widely ignored at the time because they seemed unnecessarily
complicated.
Figure 3.3. Brightness ("achromatic"), y/b and r/g opponent signals for light
throughout the spectrum. Hurvich and Jameson showed that by using a blue or
yellow AND a red or green light they could match by hue cancellation all
wavelengths of the visible spectrum, thus quantifying
the r/g and y/b components of the colour sensation induced by each
wavelength. Image source:
http://webvision.med.utah.edu/imageswv/KallColor15.jpg

To see how the colours of the spectrum can be understood as different

combinations of red vs green (r/g) and yellow vs blue (y/b) signals,
drag the yellow triangular slider in Figure 3.4 to the left.. At the long-
wavelength end, positive r/g and y/b signals occur, associated with
high L, low M, and zero S responses (Figure 3.1), and we experience a
broad zone of orange-red colours. Moving left towards shorter
wavelengths, past where the L response peaks, r/g begins dropping,
and at a wavelength around 577 nanometers passes through zero, and
we experience unique yellow, that is, a yellow that is neither reddish
nor greenish. (Note that we experience the unique hue, not where its
hue signal is at its maximum, but where the other signal is at zero). To
the left of this point negative r/g signals, broadly corresponding to a
dominant M response, result in a series of greenish colours. At around
513 nm the y/b signal drops to zero, and we experience unique green,
neither yellowish nor bluish. Throughout the remainder of the
spectrum, where S is the dominant cone response (Fig. 3.1),
the y/b signal is negative, and all of the colours are bluish in character.
As the M response continues to fall compared to the S response, we
reach a second point of zero r/g ( unique blue) at wavelengths around
475 nm, beyond which we get positive r/g signals and reddish colours
(violet) again. All of the wavelengths quoted are subject to considerable
individual variation.

Figure 3.4. Interactive demonstration of changing y/b and r/g signals

throughout the spectrum. Push the slider from the long to the short wavelength
end of the spectrum to understand how the colours of the spectrum result from
different combinations of positive and negative r/g and y/bsignals. Copyright
David Briggs and Ray Kristanto, 2007.

Not all of the 360o range of possible hues can be evoked by single
wavelengths of light. Unique red, for example, does not occur in the
visible spectrum, because all long wavelengths evoke a
positive y/b component. Colours from unique red through magenta to
red-violet can only be evoked by a mixture of wavelengths from the red
and blue/violet ends of the spectrum.

Revised 08 October 2011.

<< 1 2 3 4 5 6 7>>

ADAPTATION

We all know that the muscles of the iris can increase or decrease the
size of the pupil in response to dimmer or brighter light respectively,
but this mechanism is only a minor component of our ability to adjust
to different light intensities. Alongside the shift between rod (low light)
and cone vision, the main process involved is called adaptation, which
refers to our ability to adjust the sensitivity of the receptor cells in the
retina in response to the general level of illumination. On going into a
darkened room we may at first see little or nothing, but as the
sensitivity of our receptor cells slowly adjusts upwards, we begin to see
more and more detail. On returning to a more brightly illuminated
environment we much more rapidly (and sometimes painfully) adjust
the sensitivity of those cells to the prevailing illumination.

The importance of adaptation to colour vision comes from the fact that
the L, M and S cone systems can to a certain extent adapt
independently to the prevailing illumination if one set of cones is more
or less strongly stimulated than the others. For example under
incandescent lighting, where the S cones are less strongly stimulated
than the M or L cones, the former increase their relative sensitivity,
causing the light to seem less strongly coloured than it otherwise would
appear.

This adjustment of the input from the cones should not be (but often
is) confused with colour constancy, which relates to the processing of
that input into an interpretation as to the local colour of surfaces and
the quality of the illumination.

AFTERIMAGES AND SUCCESSIVE CONTRAST

Figure 3.5. Afterimages and successive contrast. Stare fixedly at the centre of
circle A for at least 20 seconds, then immediately look at the centre of circle B
for about ten seconds, noting the changing afterimage. Each sector will display
an afterimage the colour of the additive complement of the stimulus colour.
Repeat the procedure, this time looking immediately at the centre of circle C.
Now the colours of the afterimage in each sector will influence the appearance
of the red colour in an example of successive contrast.

The coloured afterimages seen after exposing the eye to a coloured

light for a period of time can be explained in terms of changes in the
relative adaptation of the three cone types. After a period of exposure
to coloured light, the cone type or types that are relatively weakly
stimulated by that light, due to a paucity of certain wavelengths,
become proportionately dark-adapted. When neutral light is restored, a
temporary illusion of a light composed of the "missing" wavelengths is
seen.

Ewald Hering discussed afterimages as evidence for his opponent

model of colour vision, and very numerous subsequent authors down
to the present have asserted that coloured afterimages take the colour
of the opponent colour in Hering's system. They do not - they in fact
take the colour of the additive complement - that colour of light that
mixes with the stimulus to make white light. For example, the
afterimage seen after viewing red is cyan, not green. Pridmore (2008)
examines the history of this confusion, and points out that Hering
himself never actually claimed that chromatic induction in coloured
afterimages was an opponent process, only lightness induction. The
afterimage takes the colour of the opposite stimulus, not the opposite
colour experience.

The colour of an afterimage influences the apparent colour of objects

viewed subsequently to the stimulus, a phenomenon known
as successive contrast. For example in Figure 3.5C, the afterimages of
red, yellow and magenta (the three colours that contain red light) dull
the appearance of the red, while the afterimages of the other three
colours intensifies it. Successive contrast thus also goes towards the
additive complementary.

Successive contrast resulting from adaptation is the actual explanation

of the phenomenon sometimes mislabelled "fatigue" of the eye - for
example, the apparent dulling of a high-chroma red surface after being
examined fixedly for a few seconds. In looking at a red object the L
cones are not being "fatigued" any more than they are in looking at a
white object - the effect is caused by the increased sensitivity of
the othercone types.
For some effective demonstrations of coloured afterimages see here
(link 1,2,3). The demonstration involving flashing colours (link 4)
shows that a faint afterimage is noticeable even after a very short
interval of viewing a high-chroma surface

<< 1 2 3 4 5 6 7 >>

COLOUR CONSTANCY

Our visual system does more than provide us with a visual image of the
world*. A visual field may contain an array of light and dark areas, but
which of these areas represent shadows, and which represent dark-
coloured objects? Does a red area in the visual image represent a red
object in white light, or a white object in red light? Is a particularly
bright point in the visual image a light valued surface, or a light source?
To separate the effects of lighting from object colour, the visual image
must be interpreted. Since we normally perceive object
colours without having to make a conscious effort of interpretation, it
follows that this interpretation must be occurring unconsciously as part
of the processing of the visual image by our visual system. The ability of
our visual system to automatically assign a more or less consistent local
colour to an object under illumination of differing hue, brightness, and
saturation is known as colour constancy.

Visual adaptation to the brightness and colour of the illumination,

through adjustment of the sensitivities of the three cone types, should
not be confused with this process of interpreting visual data into
illuminant and object colours, although adaptation does assist colour
constancy processing by improving the data provided by the eye on
colour differences, provided that the illumination is not too strongly
coloured or too dim. Under monochromatic light chromatic adaptation
is ineffective, so that objects can convey no information on differences
in hue and chroma to the eye, and differ only in value (see Figure
10.8, slider at right end of scale), and of course under very dim
illumination, where only the rods function, vision is also
monochromatic.

Colour constancy is never complete. Under weakly coloured

illuminants, an array of surface colours may seem to approximately
maintain their perceived hues, but changes in value in relation to each
other are inevitable (see Figure 10.8, slider near middle of scale).
Colours close to the hue of the illuminant tend to
appear relatively lighter; while dissimilar hues tend to appear darker.
In addition, two surfaces that match under a white light may not match
under a coloured light, or even under another white light with a
different spectral power distribution, a phenomenon known
as metameric failure. These factors naturally play havoc with the tonal
scheme of a painting, which is why it is recommended that paintings
should be executed under similar lighting conditions to those under
which they will be viewed.

Given the fact that it automatically interprets object colours for us, we
can surmise that our visual system must in some way be coming up
with a normally satisfactory solution to the "inverse problem" of
separating the effects of illumination and object colour throughout the
visual field. A surface giving off light whose brightness is consistent
with being a reflection of this inferred illumination is seen as having an
object colour that may be bright or more or less greyed (having a
perceived content of black). An object that seems too bright to be the
result of reflection of the inferred illuminant is seen either as
a fluorent (fluorescent-looking) object, an independent light, or a
specular reflection of an independent light. Evans (1974) proposed the
term brilliance for this scale of appearance, and used the term zero
grey point for bright colours at the point where they exhibit neither
"greyness" nor fluorence.

Several well known "optical illusions" dramatically demonstrate colour

constancy in action. In the checkerboard illusion by Edward Adelson
(Figure 3.6A), the two areas marked A and B are actually identical in
value on the image (i.e. they are the same grey), but our visual system
calculates that in a shadow area this grey must represent a white
surface, while in the lit area the same grey must represent a dark
surface, and that is how we see them. In the same way, in the cube
illusion by R. Beau Lotto (Figure 3.6B), our visual system sees the
same image colour as being a dark brown object colour in the context
of strong lighting, and a fluorent orange in a deeply shaded context. In
the cross-piece illusion , also by Lotto (Figure 3.6C), the image colour
at the intersection of the two rods is actually an identical middle grey in
both cases, but in the apparent context of a yellow translucent filter on
the left and a blue translucent filter on the right, this is judged, and
seen, to be the reflectance of a blue-grey object and a yellow object
respectively.
Figure 3.6. Three optical illusions demonstrating colour constancy in action
(follow links for larger images). A. The checkerboard illusion of Edward Adelson. B.
The cube illusion of R. Beau Lotto. C. The cross-piece illusion of R . Beau Lotto

In each case these comparisons are made unconsciously, and what we

see in our normal way of looking is the inferred local colour. Tonal
painters have to learn above all to look at their subjects with a different
attitude to normal viewing, in order to judge objectively the hue,
"colorfulness", and brightness of the light coming to from each point
in the subject. That is, we have to switch off one kind of processing -
one that is built into our visual system, wherein each colour is
compared with an inferred white - and learn a completely different
kind of processing, where each colour is compared with the full range
of colours in subject as a whole. With practice we can learn to switch at
will between our normal mode of vision and this painter's way of
seeing. But we always need to be on guard against the tendency to slip
into judging colours in constancy mode, that is, to paint their perceived
local colour, instead of the colour that we need to create the illusion of
that colour. The problem is analogous to the difficulties encountered
in foreshortening in drawing, where we also need to learn to see and
draw what is actually in front of our eyes, and not what our brain
works out for us.

At this point the beginning painter might ask: "well, if that's the way it
looks to my eyes, shouldn't I paint it that way?" The answer to this is a
definite no - if we can recreate the stimulus that created the
appearance, we will create the effect the we see in our subject; if we
instead chase the appearance, we will create something different.

Certain tricks or devices that are sometimes recommended for

observing colour can be workable, though many of these have serious
limitations. For example, the idea that you can hold up paint on a
brush, palette knife or other device and match it with your subject is in
general workable only if you have some way of turning up the
illumination on your brush until you can match the
brightest highlight on your subject with the tone of your paint, and
can keep the illumination at the same level while you compare the
other colours. (One teacher currently advertising such a method on the
internet seems to get around this problem by having his students paint
only dimly lit subjects). These methods of course eliminate the option
of translating the tonal range of the subject into a your own choice of
tonal level and range in your painting. Devices involving an aperture in
a card that bears a greyscale or colour chips for comparison suffer from
the same difficulties and limitations, and in addition run the risk of
giving an excessive impression of the brightness and "colorfulness" of
colours seen in isolation, though this can be avoided if the colours are
continually compared with the brightest colours in the subject. The
latter comparison can be made very effectively by using a blank card
with two apertures, which can be moved towards or away from the
observer in order to compare more and less separated points.

Other methods involve "squinting", "unfocussing the eyes" or using

peripheral vision, or observing the darkened and reduced image of the
subject in a flat or convex black mirror. Squinting is best understood as
closing the eyes and then just opening them enough to allow a dim
impression of the subject. It is generally said to work by eliminating
details, allowing the artist to concentrate the relationships of the big
masses of the visual field, although for myself at least, another
important factor is its effect of flattening the visual field, helping me to
view the subject in what psychologist David Katz called film mode, as
opposed to surface mode.

*I am of course ignoring here the fact that for most of us it creates two visual
images that we combine stereoscopically. Colour constancy processing is
presumably aided by stereoscopic cues, but still works effectively in distant
scenes or 2D images, where stereoscopic cues are lacking.

<< 1 2 3 4 5 6 7>>

SIMULTANEOUS CONTRAST AND ASSIMILATION

Simultaneous contrast and assimilation refer to the tendencies of the

visual appearance of a surface colour to be influenced by adjacent and
interspersed colours respectively. The causes of both of these slight
failures of colour constancy are still under active scientific
investigation. Both may be broadly regarded as reflecting the fact that
our visual system does not work like an objective measuring device,
sending raw light and colour data to the brain. We see always by
comparisons.
In simultaneous contrast the appearance of colours moves away from
that of the surrounding colour in all three dimensions of surface colour
- hue, chroma and lightness (Figure 3.7, 3.8, 3.9).

Figure 3.7. Simultaneous contrast of hue and chroma. The purplish-red colour
(A) looks more purple against red (B), more red against purple (C), lower in
chroma against bright purple-red (D), and progressively higher in chroma
against dull red-purple (E), grey (F), and green (G).

Figure 3.8. Simultaneous contrast of lightness and chroma. The same purple-
red looks darker against a lighter (B) and lighter against a darker (C) colour of
similar hue and chroma. The intensified change in appearance when combined
with a contrast in chroma (D,E) was used in the well-known optical illusion of
this form devised by Josef Albers.
Figure 3.9. Simultaneous contrast of hue. A medium grey (A) is seen against a
background of similar lightness and green (B), magenta (C), red (D), cyan (E),
blue (F) and yellow (G) hue. In each case the appearance of the colour moves
towards the hue of the additive complement.

An important special case of simultaneous contrast of hue is the

phenomenon of complementary colours in shadows. When we have a
strongly coloured main light and a neutral secondary light, the colours
in the shadows of the main light shift in appearance towards the
complementary of the colour of the light. (Do not confuse this
phenomenon with the hobby painter's recipe of adding the pigmentary
complementary of the colour of the object to get its colour in shadow).
The induced complimentary effect is the basis of the artist's
conventional rule of warm lights creating cool shadows and vice versa.

Simultaneous contrast is responsible for the well-known artists

problem that paint mixtures may look quite different on the palette
(especially a white palette) to their appearance on the painting.
Experience and a growing capacity to see colours objectively can help
here, as can devices such as mixing over a grey scale placed under a
glass palette. The strategies of using a coloured ground closer to the
tonal level of the finished painting than stark white, and of covering the
canvas at an early stage with a mosaic flat colours can both minimize
problems caused by simultaneous contrast on the canvas. The concept
of a colour space reminds us that despite the shifting appearance of
each of our colours in different surroundings, the colours in any
painting nevertheless have a fixed relationship to each other (shifts due
to different light sources excepted) - and it is the business of the
painter to get these relationships right.

Assimilation is a phenomenon having the opposite effect to

simultaneous contrast, and is seen when small areas of colour are
closely interspersed. Examples include the illusions known as the Von
Bezold spreading effect, neon colour spreading, and White's illusion. In
each case, surfaces change appearance by moving towards the colour
of the interspersed areas.

<< 1 2 3 4 5 6 7>>

3.6 "What is color?"

This page presents an explanation of colour for school children,
produced in response to the 2014 Flame Challenge by actor Alan Alda.
Whether it succeeds with its target audience or not, I hope it helps
promote a better understanding of colour, and especially of the
significance of the opponent model of colour vision, which holds that
colours are created by the eye and brain as combinations of red, yellow,
green and blue colour signals.
After reading this page you may like to watch the video version I
submitted, which can be viewed at the bottom of the page. For all
image credits please see the video or its Vimeo page.

...

My nephews Jay and Guy have been using a glass prism to study the
spectrum. Both saw a spectrum, but the colours that they saw were
very different.

Isaac Newton observed that when white light passes through a glass
prism, different rays of light are bent by different amounts, and these
rays appear different colours. Newton called this coloured band of light
a spectrum.
Newton also showed that mixing these coloured rays made white light
again.

Today we think of these coloured rays as waves of

different wavelength.
Newton also said something very strange: he said that "the rays to
speak properly are not coloured". By this he meant that a colour is
really a sensation of the mind resulting from light, just like a sound is
a sensation of the mind resulting from physical vibrations of the air.

You may know that there are three primary colours for mixing
paints. There are also three primary colours for mixing lights: red,
green and blue.
Every colour you've ever seen on your computer screen, including
white, is made by mixing lights of these three primary colours (drag
the sliders at the bottom to adjust).

PLEASE STOP AND THINK ABOUT THIS QUESTION: If the

spectrum is a continuous series of many wavelengths, why
should three colours be special? Where does that number three come
from?

AND THIS QUESTION: If the spectrum is just a series of short to long

wavelengths, why do colours form a circle?

The key to both questions is that Newton was right when he said
that the rays to speak properly are not coloured. The rays of the
spectrum differ only in wavelength; our eyes and brain add the
colours. And some of us add different colours to others.
The reason why primary colours come in sets of three is that human
eyes have three types of light-detecting cells, called cone cells, used in
seeing colour. L cones respond to all wavelengths, but respond most
to those we see as yellow. M cones also respond to all wavelengths,
but respond most to those we see as green. S cones respond most to
wavelengths we see as blue and violet.

The three primary colours for mixing coloured lights each make one
type of cone cell respond more than the other two. The three
primary colours for mixing paints each absorb one of these light-
mixing primaries. [To explain why this works is another story; if you
are interested, you should look up subtractive colour mixing].

Notice carefully that no cone cell can "detect" any particular colour
band of the spectrum, such as red or green. For example, when an L
cone cell responds to light, it doesn't "know" if it's responding to a
small amount of light from the yellow band, or a larger amount of light
from some other band.
We tell wavelengths apart by comparing how our L, M and S cone
cells respond. When we see an even range of wavelengths of light, our
L, M and S cone cells all respond similarly, and we see the light
as white.

But any single wavelength or band of the spectrum produces

an unequal response of the L, M and S cone cells. Colours
are signals that our eyes and brain create when we
detect uneven amounts of the different wavelengths of light - by
means of the unequal response of our cone cells.
Our eyes and brain create four basic colours, yellow, blue, red and
green, arranged in two pairs.

One colour signal is either yellow or blue: We see yellow in the part
of the spectrum where L and M respond most, and blue in the part
where S does.The other colour signal is either red or green: We see
red at the two ends of the spectrum, where L or S respond most, and
green in the middle, where M does.
So the colour "red" is not something we "detect" in light, but is a signal
added by the eye and brain to two parts of the spectrum that have
nothing in common physically.

You might be wondering why our eyes and brain would do this, but
look what we get when the two colour signals combine:

Colours form a circle because the combinations of the four colour

signals we create form a continuous loop: RED + YELLOW, YELLOW +
GREEN, GREEN + BLUE, BLUE + RED. Single wavelengths of the
spectrum can produce most but not all of the possible combinations.

Magenta is a combination of our blue and red colour signals,

produced exactly the same way as the colours of the spectrum, but
from mixtures of wavelengths from the two ends.

My nephew Guy's eyes have only two types of cones. So his brain can
create a yellow/blue signal, but with no middle cone, no red/green
signal, so his spectrum has only two colours.
You can actually see how much light of each wavelength an object
reflects by shining a spectrum onto it. White and grey objects reflect all
wavelengths of light about evenly, while coloured objects reflect the
wavelengths of light unevenly.

A bright yellow object like a lemon absorbs blue and violet

wavelengths, and reflects red, orange, yellow and green
wavelengths.Our L and M cones both respond strongly to these
wavelengths, and these responses:
- combine to create a strong yellow colour signal, and
- cancel each other from making a strong red or green colour signal.
[Study this section again to see if you can work out why].

On your computer screen the colour of the lemon is made by tiny red
and green lights.

Together these give off wavelengths of a similar range to the lemon

that create the same relative response of our cone cells, and so the
same colour, even though the proportions of these wavelengths are
very different.
P.S. Colour is more complex in complex scenes. A white object can
reflect light of any colour if the lighting is that colour. Our brains are
quite good at automatically allowing for these effects of lighting, and
because of this ability, called colour constancy, the colour that we see
an object as being depends not just the wavelengths coming from it,
but on the whole scene.

You probably see an object coloured like the one on the upper left in
all of these pictures, even though each of the other
five pictures actually contains a very small range of colours, from only
about a third of the colour wheel!* (click picture to enlarge).

If you study this explanation carefully you will understand much better
than most adults what colour really is. But if you take the time to
explain it to them, and they keep trying, you'll find they can eventually
understand almost anything.

...

This is the video version submitted for the competition:

You can leave comments on the video on its Vimeo page:

David Briggs - What is Colour?

...

As a test of your understanding after studying this page and the video,
you might like to watch these two YouTube videos to see if you can see
the problems with them. (I've written a technical explanation of what I
think is wrong with them here).

This Is Not Yellow (Michael Stevens, Vsauce)

https://www.youtube.com/watch?v=R3unPcJDbCc

Colour Mixing: The Mystery of Magenta (Steve Mould)

https://www.youtube.com/watch?v=iPPYGJjKVco

March 9, 2014.
<< 1 2 3 4 5 6 7 >>

*These pictures are variations on an illusion created by Dr R. Beau Lotto and made
publicly available by him for use with proper accreditation.

3.7 Answers to "What is colour?" page

The previous page presents an explanation of colour produced for
school children in response to Alan Alda's 2014 Flame Challenge, but
which may also be helpful to adults. My video attempts to explain in
simple terms the role of both trichromatic input and opponent output
in colour vision: Although the opponent model has been widely
accepted for many decades alongside trichromacy, most popular
explanations of colour vision still omit it completely.

After studying that page, and/or watching the video here, readers
should be able to spot some problems with these two rather misleading
YouTube videos, and may like to compare their conclusions with my
notes on this page.

 This Is Not Yellow (Michael Stevens, Vsauce, 2012)

 Colour Mixing: The Mystery of Magenta (Steve Mould, The
Royal Institution, 2013)
 Summary

This Is Not Yellow (Michael Stevens, Vsauce, 2012)

"Real" yellow light = "light with a wavelength of about 570

mm" (onscreen text, 0:55).

Stevens' basic point that screen yellow is made from a mixture of red
and green lights is of course correct, and judging by the comments was
news to a remarkably large number of his viewers. However, his
explanation perpetuates an incomplete and partly inaccurate picture of
the nature of colour. Like many other accounts of colour vision,
Stevens' presentation includes a version of the trichromatic theory of
input by means of three cone types, but completely omits the opponent
theory of output of yellow/blue and red/green colour signals in the
brain, and instead speaks as if colours themselves reside in the
wavelengths of the spectrum. Stevens' statement here that the
wavelengths of the spectrum that appear yellow are "real" yellow, and
his inference that yellow-looking light lacking these wavelengths is
"fake" yellow (see below), reflect the assumption that the colour yellow
'exists' in these wavelengths.
"Here in this room, this lemon is "subtractively yellow". It
absorbs all visible wavelengths of light except for yellow
light, which it reflects onto my retina."

This second mistake is a commonly repeated falsehood regarding the

reflectance of yellow objects, also found for example in Michael
Wilcox's book Blue and Yellow Don't Make Green. A lemon actually
reflects much of the red, orange, yellow and green light that falls on it,
and most of its yellow colour comes from the additive mixture of the
reddish and greenish wavelengths, like the yellow on a screen. The
same is true of all bright yellow objects:
http://www.huevaluechroma.com/045.php

"But, the screen that you are using to watch this video
doesn't produce yellow light at all; in fact, it can only
produce red, blue, or green light." ... "Absolutely no yellow
is coming off of your screen and falling onto your retina."

By "no yellow" Stevens means no light from the yellow band of the
spectrum, which is not strictly correct, because the green phosphor is
not monochromatic (below). It might also be pointed out that the
colour yellow is seen in a large part of the spectrum beyond
the pure yellow band, and that the colours of both the yellowish green
phosphor and the orangeish red phosphor have a component of yellow.
It's true however that the screen gives off considerably less light in the
pure yellow band of the spectrum than the lemon.
"The really cool, but kind of disturbing, thing about this is
that here in the room, I am actually seeing "real" yellow
light- but you are seeing "fake" yellow".

Based on his mistakes about the wavelengths reflected by yellow

objects and the external existence of yellow in the spectrum itself,
Stevens assumes that being yellow is the property of reflecting or giving
off light from only the pure yellow band of the spectrum, and describes
the computer screen as "fake yellow" because it does not meet this
criterion. But given that light we see as yellow can contain either a
single band of the spectrum (spectral yellow) or mostly two bands
(screen yellow), or more than half the wavelengths of the
spectrum (the light reflected by all bright yellow objects including
lemons), it is impossible to justify Stevens' criterion. Since most of the
yellow light we see in nature is a mixture of red to green wavelengths,
one could actually argue that this broadband yellow is "real "yellow,
and that the single-wavelength yellow of the spectrum is "fake" yellow.
However, instead of talking in terms of real and fake yellow, it makes
far more sense to say that yellow is a mental perception, not a
wavelength.

Accepting this viewpoint, the title sentence "This Is Not Yellow" is

actually true of the screen (and also of the lemon) in the specific sense
that the colour yellow does not exist outside the observer, in the same
way as Newton said that the rays to speak properly are not coloured.
However in our normal, anthropocentric manner of speaking we would
say that both the screen (unless magnified) and the lemon are yellow,
because they are perceived to be yellow. This choice of
alternative semantics should not be confused with the actual question
of whether colour itself does in fact reside in the mind (as the modern
scientific explanation indicates) or externally in the world (as some
academic philosophers still maintain).

Stevens would have been perfectly correct if he had simply said that the
screen gives off a very different distribution of wavelengths to the
lemon, and that our visual system only detects what is called
the dominant wavelength of light, not the actual wavelengths present.

"Our retinas contain three types of cone cells that are

receptive to colour, and each one is best suited to detect a
certain colour. One is great for blue, the other is great for
green and the third is great for red. Notice that there is no
individual cell looking for yellow."

The modern scientific explanation of colour vision, as set out on the

preceding page, combines the trichromatic (three cone) model with the
opponent model of hue perception. Like many other popular
explanations of colour, Stevens describes the trichromatic input of the
three cone types, but makes no mention of opponency, and in fact
offers no explanation of where colours themselves actually come from.
This omission, and his choice of words here, create the impression that
colours exist in light irself and are 'detected' by the cones.

Apart from the wording that cone cells respond to colours rather
than wavelengths, this part is a fair description of the original 19th
century version of the trichromatic model of colour vision. We have
since discovered that the long-wavelength (L) cone is
actually most sensitive to the yellow band of the spectrum, a fact that is
even shown in one of the diagrams that Stevens displays. However it
remains true that a blue, green and red light will respectively cause the
S, M and L cones to each respond more than the other two.
"All a computer monitor or mobile phone screen has to do to
make you think you're seeing yellow is send a little bit of
red and a little bit of green light at you. As long as the pixels
and the little sub pixels on them are small enough that you
can't distinguish them individually then your brain will just
say well I'm receiving some red and some green, that's what
yellow things do, hmm, it must be yellow, even though it
actually is not."

Stevens presents here a description of colour vision in which colours

exist physically in the spectrum, and the brain identifies colours based
on the colours it supposedly detects with individual cone cell types. He
assumes that the perception of yellow is an attempt to identify "real"
yellow, meaning the wavelengths of the spectrum we see as pure
yellow, and therefore considers the brain to be "lied to" when it sees
other combinations of wavelengths as yellow.

The modern scientific view of colour vision instead explains how the
spectrum of colours is a creation of the eye and brain (see "Summary"
below). The brain can not "say" it's receiving some red and some green,
becuase no cone cell can "detect" any particular colour band of the
spectrum. The yellow colour signal is not a guess that light is from the
narrow band of wavelengths we see as pure yellow, but is the response
of the visual system to all middle- to long-wavelength light. A
balanced response of the L and M cones produces pure yellow because
the red/green colour signals cancel out.
Colour Mixing: The Mystery of Magenta (Steve Mould, The
Royal Institution, 2013)

"Purple is a weird colour" ... "your brain invents a colour, it

makes up a colour, and that colour is magenta"

Mould is perfectly correct in saying that the brain "makes up" magenta;
the problem is that his statement leaves the impression that the colours
of the spectrum are not 'made up', and so must 'exist' in light itself. The
notion that magenta alone is not "real" has never been accepted
science, but is encountered elsewhere on the internet (see Michael
Moyer's Stop This Absurd War on the Color Pink). The idea is
apparently based on the assumption that the spectral colours are real
because they "exist" in individual wavelengths of light. Magenta lacks
these credentials and so is said to be "made up" by the brain when it
looks at mixtures of wavelengths from the two ends of the spectrum.
This argument itself seems a bit "weird" because magenta gives every
appearance of being simply a mixture of the red and blue or violet
colours supposed, on this assumption, to "exist" in those light
mixtures. However magenta actually is "made up" by the brain, but so
are all of the spectral colours as well. In the modern theory of colour
vision, all colours are created by the brain in exactly the same way, as
combinations of yellow/blue and red/green colour-opponent signals
based indirectly on unequal responses of the cone cells, regardless of
whether the stimulus is a single wavelength or a mixture.
Mould's essential point here is that the colour of a mixture of
wavelengths from the ends of the spectrum is not the colour of the
average of these wavelengths (green), but is an entirely different
colour. This could be less problematically expressed by saying that
colour is the mental perception of what is called dominant wavelength,
which is not a simple the average of the wavelengths present, but has a
360 degree range, such that balanced mixtures of all wavelengths have
no dominant wavelength, and appear white. (If dominant wavelength
was a simple average, balanced mixtures of wavelengths would also
appear green). The reason why this happens rests on the very factor
that is missing from Mould's explanation - the opponent processing of
cone responses, beginning in the retina.

"... you have these cone cells at the back of your eyes that
are sensitive to different parts of the spectrum, so when red
light comes into your eyes there's a set of cones that fire and
tell your brain you're looking at something red, so we'd call
those the red cones. There's another set of cones that are
more sensitive to green, so when there's green light going
into your eyes they fire and they send a message to your
brain, and there's blue cones as well. So you've got red
cones, green cones, and blue cones. So what about yellow?
What about when you're looking at yellow light, like that?
Well in that situation, you don't have a yellow cone. So
what do you do? Well, yellow is quite close to red so your
red cone fires a bit, and yellow is quite close to green as well
so your green cone fires a bit ... So your brain is getting a
message from your red cone and your green cone at the
same time and it's deciding OK well I must be looking at
something in between those two colours, and that's brilliant
because your brain is perceiving something about the world
that it isn't able to measure directly; it isn't directly
sensitive to yellow light."

This part describes the 19th century version of the trichromatic model
of colour vision, before it was discovered that the three cones have
surprisingly broadly overlapping sensitivities, that the L cone actually
responds most to the yellow band of the spectrum, and that the brain
does not get "messages" directly from the three cones, but creates
colour signals based indirectly on impulses from the retina that
convey differences between pairs of cone respones. Like Stevens,
Mould makes no mention of opponency.

It can be seen from the excerpt as a whole that the words "isn't directly
sensitive to yellow light" are just very carelessly chosen, and that
Mould is aware that the so-called "red" and "green" cones do in fact
respond to light from the yellow band of the spectrum (though
apparently not that the so-called "red" cone actually responds most to
the yellow band). These words however gives the highly misleading
impression that the cones each respond to a particular colour band of
the spectrum, and don't respond at all to other bands. Indeed they even
seem to have misled the writer of the caption to this video on its
YouTube page, who says (completely incorrectly): "The cone cells
within our eyes ... are only sensitive to Red, Green and Blue light. So
how are we able to see so many colours when we can only directly
detect three...". This particular misconception now seems quite
common, and may stem directly from this source.

One could avoid introducing nonsense about cones being sensitive to

red and green but not to yellow light simply by saying that
the relative response of the L and M cones changes progressively
through the long-wavelength part of the spectrum, and that this
progressive change in the difference between the L and M cone
reponses is experienced as the colours red, orange, yellow, yellow-
green, and green.
"It does mean that you can be tricked. ... there's no yellow
light here, but you'll see yellow anyway"

Like Stevens, Mould means by this that wavelengths from the yellow
band of the spectrum are entirely absent from his mixture of red and
green lights, which is probably not strictly correct, but the actual issue
here is the tacit assumption that seeing yellow is an attempt to identify
a particular narrow band of wavelengths of the spectrum, and hence
that we are "tricked" when we see a mixture of red and green
wavelengths as yellow. Once again, yellow is a perception, not a
wavelength, and both yellows are equally real perceptions. It would
be more accurate to say that we are all tricked into assuming that
yellow is a physical property residing in light, when really it is a
mental perception caused by many physically different lights that
happen to have the same effect on our cone cells.

Summary

As proposed by Young and Helmholtz in the 19th century, the

trichromatic theory held that three receptors sensitive to the parts of
the spectrum seen as red, green and blue or violet
generated fundamental sensations, and explained the sequence of
spectral colours as successive combinations of these three sensations in
changing proportions. Yellow was considered to be one such
"compound" sensation composed of red and green sensations. In the
modern zone theory, an updated form of the trichromatic receptor
model is coupled with the opponent model of colour perception. The L,
M, and S cones respond most strongly to the parts of the spectrum seen
as yellow, green and blue-violet respectively, and the brain creates
yellow/blue and red/green opponent colour signals based indirectly on
signals representing differences in the cone responses. The sequence of
colours of the spectrum is explained as the product of successive
combinations of these red, yellow, green and blue opponent hue
signals.

The two YouTube videos incorporate the

receptor sensitivities envisioned in the 19th century version of the
trichromatic model, but make no attempt to explain the sequence of
colours of the spectrum as combinations of either three "fundamental
sensations" or four opponent colours. In fact they give no indication at
all that these colours originate in the brain. Instead they couple the
early trichromatic receptor sensitivities with explanations that speak of
the spectral colours as if they are physically real and reside in light
itself. The omission of any explanation of the sequence of colours in the
spectrum adds to the impression that this sequence simply exists there.

This combination of premises leads both Stevens and Mould to offer an

inverted explanation of the perception of yellow that bears little
resemblance to present or past science. Real yellow is deemed to reside
in the spectrum in the wavelengths we see as pure yellow. Although
there is supposedly no cone that is "looking for" (Stevens)/"directly
sensitive to" (Mould) these pure yellow wavelengths, the brain
somehow knows of the existence of this "real" yellow in the spectrum,
and its location between red and green. The perception of yellow is
treated as a process of identification of this invisible, externally
existing yellow by a defective criterion that can result
in misidentification. This criterion is the balanced response of the so-
called red- and green-detecting cones.

In the modern scientific explanation the L cone (their "red" cone) is

known to be most sensitive to the yellow band of the spectrum, but
much more to the point, the L cone is not a yellow 'detector' or a red
'detector'. Instead, the L and M cones detect essentially all wavelengths
of visible light, and high L and M cone responses both contribute
to creating a yellow colour signal throughout the middle and long
wavelength parts of the spectrum. Through most of this range this
yellow signal is combined with either a red signal (where the L cone
responds most) or a green signal (where the M cone responds
most). Pure yellow is seen when the L and M responses are balanced
and the red/green signal cancels out. This occurs in the spectrum in a
narrow range of wavelengths near 570 to 580 nanometres, but also for
exactly the same reason in any mixture of wavelengths that has the
same relative effect on the L and M cones.

It's likely that Stevens and Mould were not especially concerned with
colour itself, and mainly intended to make the valid point that
physically the brain effectively determines the dominant wavelength of
light by means of the relative response of the three cone types. It's just
unfortunate that in doing so they create or perpetuate misconceptions
about the fundamental nature of colour, the details of the colour vision
process and (in the case of Stevens) the physical basis of object colours.
The combination of outdated cone physiology and colour "realism" that
these videos reinforce constitutes perhaps the most widely-held view of
colour vision among non-scientists today. Although the details of how
the colour-opponent signals are generated within the brain are still
mysterious, the opponent model itself has been widely accepted in
science for many decades. In selecting the topic "What is color?", the
2014 Flame Challenge has certainly highlighted an area where
scientists have largely failed to convey a modern understanding of the
topic to a wide audience, and apparently even to some science
communicators.

March 9, 2014.

<< 1 2 3 4 5 6 7 >>

PART 4. ADDITIVE MIXING

4.1 ADDITIVE PRIMARIES
  Introduction: does additive mixing matter to painters?
 Do additive primary colours exist?
 Different additive primaries for different methods of
mixing

Introduction: does additive mixing matter to painters?

By beginning this account of colour mixing with the topic of additive

mixing, I have already aligned the approach taken here with modern as
opposed to "traditional" colour theory. In the decades following the
mid-19th century revolution in our understanding of colour, most
scientists accepted the Young-Helmholtz model in its original form,
which held that colour perceptions were mixtures of the "fundamental
sensations" red, green and violet/blue. Understandably, these three
came to be regarded as the "true" primary colours, displacing red,
yellow and blue, which had until then been widely regarded as the
three fundamental colour sensations and even the three physical
components of white light. Those writers on painting who continued to
base colour theory on the older primaries often felt obliged to justify
their course in the face of current scientific opinion, and many did so
with an argument to the effect that the mixing of light was irrelevant to
painters, who must work with paint where the old primaries still
applied. Similar statements have continued to appear in "traditional"
colour theory down to the present.

This argument has lost relevance in that many painters now paint with
light (i.e. digitally), but for many reasons it was also never valid in
relation to physical media. To begin with, the colours of artists'
paints (Section 4.5) are the result of additive mixing of the wavelengths
the paints reflect. (Many "traditional" colour theory writers, including
Wilcox and Feisner, miss this point and completely misrepresent the
physical origin of colour when they endeavour to explain it). Second,
when paint colours combine optically (Section 4.4), for example when
interspersed as fine dots, the result follows the rules of a kind of
additive mixing called additive-averaging, even though physical paints
are involved. And even when paints are physically mixed (Section 6.1) ,
although the result depends to a large degree on subtractive mixing, it
generally also involves a significant component of additive(-averaging)
mixing. In any case, subtractive mixing (Section 5) can not be correctly
explained until additive mixing is understood, and it was the discovery
of the connection between these two that led to the realization that the
ideal primaries for colourant mixing are not in fact yellow, red and
blue, but yellow, magenta and cyan. Finally, since paintings are
experienced by the light reflected from them, there is a strong
argument that additive complementaries (Section 4.3) are relevant to
"colour harmony" and the visual impact of paintings.

Do additive primary colours exist?

We saw earlier that the three cone types effectively divide the visible
spectrum into three bands, in each of which the response of one cone
type predominates over the other two, so that by mixing lights from
each of these three bands we can produce light stimuli invoking strong
cone-opponent signals, and thus strong colour signals, throughout the
360 degree range of possible combinations. This is how the colours on
your computer screen are generated, from phosphors of just three
colours, a red or orange-red (R), a yellowish green (G), and a blue or
violet-blue (B). All of the colours that you have ever seen on a normal
computer or television screen, including white and grey, were created
by mixtures of lights of three such colours (Fig. 4.1.1).

Figure 4.1.1. Generation of screen colours from different combinations of RGB

lights.

Any three coloured lights have a range or gamut of colours that they
can produce by mixing, and are unable to mix colours outside this
gamut. In common speech this process of "mixing colour" by
combining coloured lights is called additive mixing, and the hues of
the lights that yield the largest gamut of additive mixtures are called
the additive primary colours. This usage of the term "primary colours"
is in accord with the original meaning of the term, as introduced in
English by Robert Boyle, for the general hue categories of physical
paints and dyes that yield a surprsingly large gamut of mixtures,
though not all hues at their maximum chroma. This usage should not
be confused with another usage of the term "primaries" found in
scientific and technical literature, for any specific lights used in
additive mixing, and even for mathematical combinations of positive
and negative quantities of such lights that do not correspond to any
physically possible light.

It is a mistake to equate the additive primary colours with the hues of

any specific R,G and B primaries, but it is equally mistaken to overstate
the arbitrariness of these hues. Although the precise hues are flexible,
to be effective as additive primaries the three stimuli must remain
within the basic hue categories of red to orange-red, yellow-green to
green, and blue to violet respectively, depending in part on the exact
kind of additive mixing involved (see below). The additive primary
colours are not perfect, in the sense that we could mix a full range of
hues at maximum saturation, but neither are they arbitrary; they
are optimal. If we instead were to define primary colours as colours
from which we can mix all other colours (an assumption of most
"traditional" colour theory), then it would of course be true that
additive primary colours do not exist, as Bruce MacEvoy has argued. In
short, additive primary colours are real if we define that term
realistically, and not real if we do not.

It is tempting but fallacious to slip from the observation that light of a

wide range of colours can be made from lights of three colours, to the
assumption that it is the colours themselves that are mixing, that the
colours obtained are therefore made of the colours of the component
lights, and that the additive primary colours are thus the ultimate
components of all colours. It's true that for some combinations of lights
this assumption (that additive mixing is literally colour mixing)
appears to hold: R and B lights mix to make a colour that is reddish
and bluish (magenta), and G and B lights make a colour that is
greenish and bluish (cyan). But mixing R and G lights can make a pure
yellow colour that isn't reddish or greenish! This apparently anomalous
failure of the colour of the mixture to contain the colours of its
components is one reason why this combination of lights is always so
surprising. (The stark difference from the behaviour of paint mixtures
is another). In its original form the Young-Helmholtz trichromatic
model assumed that all colour sensations are indeed composed of red,
green and violet fundamental sensations associated with the three
retinal receptors, and thus that yellow is a "compound sensation" made
of red and green fundamental sensations, and white is a compound
sensation made of red, green and violet fundamental sensations. This
view is still widely held among the general public today, in part because
most popular science explanations of colour vision explain the
trichromatic model of sampling by three cone types, but completely
ignore the opponent model, which is an equally important component
of the modern zone theory of colour vision.

According to the zone theory, every coloured light has a disposition to

create cone-opponent responses in the retina that are indirectly
translated into red/green and yellow/blue colour signals experienced
in the mind. The colour yellow is an independent opponent colour
created by the visual system, and does not "contain" the colours red
and green. We can loosely think of additive mixing as
a vector addition in which these colour-invoking dispositions either
add together or cancel out. For example, in mixing R (orange-red)
and G (yellowish-green) lights, the red- and green-invoking
components cancel out, while the yellow-invoking components in both
colours reinforce, although as you may notice the resulting yellow light
is somewhat less saturated than either of its component lights.
Similarly, although white light from a computer screen is a mixture of
physical components that separately appear red, green and blue, the
white colour we perceive is not a mixture of colours; rather, the
red/green- and yellow/blue-invoking dispositions of the three
component lights cancel out to leave the light mixture colourless.

Different additive primaries for different methods of mixing

Which precise hues are optimal as additive primaries depends on the

nature and purpose of the device doing the mixing:

1. For additive mixing of monochromatic lights: spectral red, green and

violet.
2. For efficient additive mixing of lights on a monitor or TV screen: red
or orange-red, yellowish green and violet-blue or blue
3. For additive-averaging mixing of object colours, as in spinning-disc
mixture (ideal): optimal orange-red, green and violet-blue

For mixing monochromatic lights (lights of a single wavelength),

the approximate maximum gamut of natural object colours is said to be
generated by lights that are extreme spectral violet, slightly yellowish
green and extreme spectral red. Fig. 4.1.2 shows this gamut on a CIE
xyY or chromaticity diagram. On this diagram the horseshoe-shaped
line represents the monochromatic lights of the spectrum, the straight
line represents mixtures of extreme spectral red and violet, and
saturation increases outwards from a "white point" to these absolute
limits. A property of the CIE chromaticity diagram is that additive
mixing paths follow straight lines, so the gamut forms a straight-sided
triangle. It will be evident that no combination of three real lights,
which must lie within the rounded horseshoe shape, can have a gamut
that encloses the entire shape.

Figure 4.1.2. CIE 1931 chromaticity diagram showing the approximate

maximum gamut of natural object colours obtainable with three monochromatic
lights (Hardy and Wurzburg, 1937), a representative RGB gamut called sRGB,
and colours of the three primaries on one of James Clerk Maxwell's 19th century
spinning disc devices. (Maxwell's pigments, particularly the vermilion, may
have dulled with time). Beside actual pigments, the theoretical optimal colours
would have a fluorent appearance, much like the RGB primaries in the second
version of the disc. Spectra of CRT screen primaries are from Wikipedia.

400 nm and 700 nm lights have low brightness for their physical
energy, so using these monochromatic lights would be a very energy-
inefficient way of creating colour on a computer screen. The
primaries used on monitors are chosen as a compromise between an
acceptably broad gamut on the one hand and energy efficiency on the
other. The compromise that is chosen varies greatly among different
manufacturers, particularly for laptop screens where energy efficiency
directly impacts on battery life. Figure 4.1.2 shows a standard colour
space called sRGB as a representative RGB gamut. Laptop screens may
have a smaller or larger gamut than sRGB.

Surfaces that reflect monochromatic light would be extremely dark in

appearance and thus completely ineffective for additive processes such
as spinning discs, which involve averaging of light stimuli over an
area. Here the biggest gamuts in theory would be obtained from
orange-red, yellow-green, and violet-blue optimal colours, each
completely reflecting about a third of the spectrum, and completely
absorbing the rest. In the mid-nineteenth century James Clerk Maxwell
approximated these colours in his spinning disc experiments using the
pigments vermilion, emerald green and ultramarine. Beside actual
pigments the theoretical optimal colour primaries would be
extremely high-chroma and fluorent (fluorescent-looking), closely
resembling the RGB screen primaries in the second disc in Fig. 4.1.2.

Page modified August 5, 2012 and July 23, 2014. Original text here.

<< 1 2 3 4 5>>

4.2 ADDITIVE MIXTURES

Let's now look in detail at the process of additive mixing on a computer

monitor (Fig. 4.2.1).
 Interactive demonstration of additive mixing of RGB screen colours.
Figure 4.2.1.
Sliders control the brightnesses of the R (top), G(middle) and B (bottom) screen
phosphors on a perceptual (nonlinear) scale, like the RGB values seen in
Photoshop. Copyright David Briggs and Ray Kristanto, 2007.

With all three lights at maximum intensity the result is white light. This
means that in terms of our opponent processing, the redness vs
greenness (r/g) and yellowness vs blueness (y/b) opponent signals are
both zero. Additionally, in terms of our lightness perception,
the screen is seen as having a greyscale value of absolute white,
because based on the complete array of colours seen, we judge the
brightness of the result to be the maximum possible. With all three
RGB lights at 50% on their (perceptual) brightness scales,
the light coming from the screen is still white, but the screen is seen
as middle grey.

Whenever the RGB primaries are unequal in brightness, however,

the r/g and/or y/b signals are unbalanced, and the screen appears
coloured. The most saturated coloured mixtures are obtained by
combining the red (R), green (G) and blue (B) primaries two at a time,
and having the third primary at zero. When R and G are both at their
full intensity with B on zero, we see the result as bright yellow.
Reducing one or other component shifts the yellow towards red and
green respectively. The mixing of red and green light is surprising
visually because it involves generation of a strong yellow signal
(strongly positive y/b) only subtly evident in the components,
and cancellation of their conspicuous r and g signals. (It is additionally
unexpected to anyone used to the results of mixing paints). Less
surprisingly, mixing the B and G lights in various proportions creates a
continuous range of visually intermediate hues with generally
negative y/b and negative r/g, passing through a blue-green colour
called cyan, while R and B in various proportions generate a range of
hues of negative y/b and positive r/g, passing through the red-violet
colour we call magenta. Together these pair-wise combinations of
primaries span the entire hue circle (Fig. 4.2.2). "Monitor" yellow,
magenta and cyan are all lighter than the primaries forming them, for
the obvious reason that they are additive mixtures of both of those
primaries.
Figure 4.2.2. Additive mixing of RGB colours.

Less saturated (more whitish) colours are produced by adding the

third primary. You will see that the position of the slider for the third
primary in effect controls the proportion of white light (left from the
slider) to coloured light (right from the slider) in the mixture. The
parameter of saturation (S) in HSB space is a measure of the coloured
component of an RGB colour, and increases as the smallest of the three
RGB components decreases.

Keeping the ratio of R/G/B constant while varying the absolute

brightnesses generates a series of mixtures of uniform hue and
saturation. The maximum brightness of this series is reached when one
of the RGB components reaches 100%, because no greater brightness is
possible without changing the R/G/B ratio. The position of the slider
for the highest primary in effect determines the proportion of black
(right from the slider) in the colour. The parameter of brightness (B) in
HSB space a measure of the brightness of an RGB colour relative to the
maximum possible at that hue and saturation, and increases as the
largest RGB component increases.
Mixing of projected light beams (Fig. 4.2.3) is another example of
additive mixing, and produces the same results as we saw with on-
screen additive mixing, with one important exception: the mixtures are
seen as light rather than object colours, and so the less bright mixtures
are seen as dimmer light instead of greyed object colours. (You will
however see greyed colours if you look at Figure 4.2.3 as a 2D image
instead of a 3D scene, and concentrate on the image colours present).

Interactive demonstration of additive mixing of spotlights. Top, middle

Figure 4.2.3:
and bottom sliders control the brightnesses of the red, green and blue spotlights
respectively. Copyright David Briggs and Ray Kristanto, 2007.

Looking back to Figure 4.1.2 you will see that the light from the screen
that we experience as white has a spectrum made up of uneven spikes.
This is possible because these spikes are distributed through the red,
green and blue-violet bands of the spectrum in such a way as to cancel
each other out, producing r/g and y/b opponent signals that are both
zero. This will also occur if all wavelengths of the spectrum are
represented in equal energy, as is approximately the case with daylight,
and also occurs with the even spikier spectral distributions of
fluorescent lighting, as long as the inputs from the three bands of the
spectrum balance out. The phenomenon of different spectral
distributions looking identical in colour, i.e. being indistinguishable to
our visual system, is called metamerism. Metameric differences in the
spectra of similar-looking light sources can cause potentially serious
annoyance for painters, in that they can cause substantial shifts in
tonal and colour relationships within a painting.

Page modified August 5, 2012. Original text here.

<< 1 2 3 4 5 >>

4.3 ADDITIVE COMPLEMENTARIES

We've seen that each of the three colours yellow, magenta and cyan can
be mixed from two of our additive primaries, and also that the three
primaries make white light. It follows that yellow, magenta and cyan
can each be thought of as being the complementary
colour or complement of the one primary that is absent from the
mixture, i.e. as needing the addition of that primary to make white.
Magenta lacks green, cyan lacks red, and yellow lacks blue.
Figure 4.3.1. Additive complementary relationships.

The third of these complementary relationships is the most surprising

one for painters, who know that if they mix yellow and deep blue
(ultramarine) paints they will get dull green. The status of cyan instead
of green as the additive complement of red, and of magenta instead of
red as the additive complement of green, may also be surprising to
artists who have internalized the conventional complementaries of the
historical "artists' colour wheel". We will see however that actual paint
mixing complementaries align more closely with these additive pairs
than with the conventional "colour-wheel" complementaries.

Since all of our possible RGB colours can be mixed from three
primaries, we can sum up their additive mixing relationships on a
triangular diagram, with the primaries at the corners and the relevant
mixed colours in the middle of each side (Fig. 4.3.2). This arrangement
automatically places each pair of additive complementaries opposite
each other.
Figure 4.3.2. Summary of additive mixing relationships of RGB colours.

Digital painters will recognize that each of these pairs of additive

complements lie opposite each other (i.e. separated by 180 degrees) on
the HSB hue circle used in Photoshop and other programs. They might
well assume that all opposite hues in the HSB circle are
complementary, and indeed, all opposite hues behave as if they are
exact complementaries when combined in layer modes such
as Normal (at varying opacity), Screen and Linear Dodge. However if
light of these "opposite" hues is combined physically instead of by
software, it is found that away from the yellow-blue, red-cyan and
green-magenta axes, opposite hues are not true
complementaries (Fig. 4.3.3). This is because the hue angle (H)
formula uses nonlinear RGB values, whereas truly complementary
RGB colours must have linear RGB values (proportional to physical
light energy) that add up to make white light. (Nonlinear RGB fits more
gradations into the lower end of the light intensity scale, where a given
change in physical energy has a greater effect on perceived brightness).
The "opposite" hues behave as if they were complementary when
combined in additive layer-mixing modes because the layer-mixing
formulae also (unfortunately) operate on nonlinear RGB values1.
Figure 4.3.3.
Comparison of optical mixing of "opposite" HSB hues and true
complementaries. Provided your monitor operates on a standard gamma of 2.2,
the two lined squares should look dull magenta and grey respectively. (The
image must not be resized, and may work best viewed at ninety degrees to the
screen and from a distance of two metres). The greyed appearance of the
squares is typical of additive-averaging and opposed to simple additive mixing
processes (see next page).

Since hue angle can not be relied on, it may be useful to provide a
diagram displaying a series of accurate complementary pairs (Fig.
4.3.4). In addition to showing true complementaries, this linear hue
circle is more evenly spaced perceptually than the standard HSB hue
circle, in which hues change slowly near red, green and blue, and much
more rapidly near yellow, magenta and cyan. (The linear circle slightly
overcompensates for this defect). The hues in Fig. 4.3.4 can not be
precisely related to absolute hue scales such as the Munsell system,
because they will vary on different monitors having different RGB
phosphors. Opposite hues will nevertheless remain true
complementaries as long as a standard gamma of around 2.2 is in
operation. Thus on a monitor in which "Monitor blue" is relatively
purplish, "Monitor yellow" will be relatively greenish.

Figure 4.3.4.
Circle of true complementaries, showing 24 HSB hues spaced evenly
according to linear RGB, opposite their true additive complementaries, and
contrasted with that complementary as tint and shading series (inner circles).
"Process cyan" and "process magenta" should be taken as broadly representative
of the actual colourants employed for those hues, as distinct from the "monitor"
(RGB) versions of those names, which are distinctly greener and purpler
respectively.

1The problem of getting the layer mixing formulae to operate on the linear RGB
values, and thus to accuately emulate light mixing behaviour, can be solved by
setting "Blend RGB values according to Gamma: 1.0" under Edit>Color settings. I'd like to thank
abstract painter Andrew Werth for suggesting this solution when I raised the
problem on the Rational Painting forum.

Page modified August 5, 2012. Original text here.

<< 1 2 3 4 5>>

4.4 ADDITIVE-AVERAGING MIXING

When two or more light stimuli can not be separately distinguished,

the light from those stimuli is seen as a single colour that follows the
same rules of additive mixing as established for physically mixed lights.
This type of colour mixing has long been known as optical mixing, a
term that refers to the idea that the coloured light mixes "in the eye" of
the observer. This should not be taken to imply that the process is more
subjective than other kinds of stimulus mixing; such effects can be
photographed with a camera as well as seen first hand. When the
stimuli can not be distinguished because they are too small, the process
is spatial optical mixing, as for example of the RGB subpixels on a
computer screen, or finely interspersed dots of ink in a printed
image.When the stimuli can not be distinguished because they are
moving too fast, the process is temporal optical mixing, for example of
the colours of a spinning disc.

Although the colours of the light stimulus follow the rules of additive
mixing, most examples of optical mixing differ from the physical
mixing of lights in that the light stimulus is seen as a property of
an object rather than an independent light, and is therefore perceived
as an object colour, judged relative to white object (see Fig. 1.1.4).

In the example of the RGB subpixels on a screen, where the light

is emitted from the object, we judge the colour mixtures in relation to
a white composed of all three RGB lights at their maximum brightness.
This means that colours such as grey, brown, olive etc, that are not
seen as colours of independent lights, are seen if the RGB lights are at
low intensity, while white and bright colours can be seen if the lights
are at full intensity. However in the examples where light
is reflected from objects, i.e. printed or painted dots and spinning
discs, the additive mixing is of a particular kind known as averaging
colour-stimulus synthesis (Burnham et al., 1967), or more
simply averaging or additive-averaging mixing. Additive-averaging
mixtures, viewed normally, can not appear as white or bright colours,
because the reflected light is in effect averaged over the area of the
object instead of simply being added. For example in Figure 4.4.1, very
little violet-blue light is reflected from the area of the yellow disc, and
little red, yellow or green light is reflected friom the area of the
ultramarine disc, so while the light reflected from the spinning discs is
white, the amount of white light is less than a white disc would reflect,
and the colour of the disc is therefore seen as grey.

Figure 4.4.1. Spinning disc displaying additive-averaging mixing of additive

complementaries, Ultramarine and Cadmium Yellow Light. These pigments
would mix physically to produce a green rather than a grey mixture.

Additive-averaging mixing can also be demonstrated on a light-

emitting monitor if the mixing involves averaging of discrete areas of
colour that are interspersed rather than superimposed (Figs 4.4.2,
4.4.3). Depending on the viewing distance, as the stripes become
narrower, the light from them is eventually seen as a single colour
whose hue and saturation follow the laws of additive mixing, but
whose brightness is less than would result from simple additive
mixing. Thus in Figure 4.4.2, where each pair of colours are additive
complementaries, the resulting mixtures are neutral, because the total
numbers of R,G and B phosphors glowing are equal in each case.
However, because only half the number of phosphors are glowing as in
an area of white screen, the mixture is seen as grey, not white. (The
nonlinear response of our visual system to brightness explains why the
grey looks more than half as bright as white). Similarly in Figure 4.4.3,
the mixtures have the hue and saturation but not the brightness (and
thus chroma) of the colours that would be expected from simple
additive mixing.

Figure 4.4.2. Additive-averaging mixing of additive complementaries.

Figure 4.4.3. Additive-averaging mixing of additive primaries.

The term partitive mixing is sometimes used in the broad sense of
optical mixing, and sometimes in the narrower sense of additive-
averaging mixing, as used here. Additive-averaging mixing, combined
with subtractive mixing, plays an important role in physical mixtures of
opaque paints (see Part 6).

There is a widespread myth in art teaching that optical mixing of paints

can produce brighter and/or stronger colours than can be produced by
the physical mixing, and or at least that it was used with this misguided
intention by the Neoimpressionist painters beginning with Seurat. An
optical mixture of two or more paints is certainly higher in value than a
physical mixture of the same paints, but optical mixtures as a whole are
not lighter or more chromatic than physical mixtures. Indeed, optical
mixtures of varied paint colours tend, due to the averaging principle, to
lie towards the middle of colour space, i.e. medium value and low to
medium chroma. The pointillist technique of the Neo-Impressionist
painters utilized optical mixing that, because of the size of the dots, was
only partial at the intended viewing distance, to produce a "soft"
yet lively effect not obtainable using physical paint mixtures, based on
the recommendation of Ogden Rood in his 1879 book Modern
Chromatics. The myths mentioned above seem to derive from a
misreading of Rood's book by contemporary and later critics.

Page modified August 5, 2012. Original text here.

<< 1 2 3 4 5>>

4.5 ADDITIVE MIXING AND OBJECT COLOURS

We all know that additive mixing applies to colours of lights, but it is

also intimately involved in the colours of objects, including the
"colours" we paint with in traditional as well as digital media. In terms
of physical stimulus, our digital colours are additive mixtures of highly
saturated red, green and blue lights, and in the same sense, paint and
other object colours are additive mixtures of the spectral wavelengths
that the object is disposed to reflect or transmit. In each case the
combination of wavelengths in the mixture determines the hue, value
and chroma of the colour.
To have high chroma, the light remitted by an object must combine
high saturation with a high level of brightness (section 1). Digital
colours are simpler than paint colours in that for each hue there is a
specific colour that combines the maximum possible HSB saturation
(S= 100) and brightness (B= 100), and can therefore be called the
highest chroma or "full-chroma" representative of that hue.
We've seen that these two conditions respectively require that one of
the RGB components is at zero while another is at the maximum (i.e.
255); the third component can be anywhere in between. In contrast,
low-chroma digital colours are additive mixtures of three RGB
components that are all similar in relative brightness.

Among physical paints the highest-chroma representatives of each hue

may seem to form a set of colours that are uniformly "pure", or "fully
saturated" in the loose sense of the word. In reality they are a
miscellaneous collection of substances united only by the fact that
the combination of saturation and brightness of their reflectances
gives the highest chroma known for their hue among acceptably
lightfast materials. These substances all fall short of the maximum
chroma that is theoretically possible, and much more so for blues,
greens and purples than for reds, oranges and yellows. If these
substances look uniformly pure in colour, it is because they are all at
the limits of the range of chromas we have learned to expect for each
hue from our experience of the world. Materials beyond these usual
limits tend to appear fluorent (fluorescent-looking) to us: they include
certain intensely coloured dyes, as well as materials that are actually
physically fluorescent.
Figure 4.5.1. Reflectances of high-chroma red, orange and yellow paints at full
strength and as 2% tints with white. The reflectance graphs in these figures
(from Mayer, 1991) illustrate typical curves for each pigment, and are not based
on the specific images (from the Dick Blick website) used here to illustrate each
paint.

High-chroma red pigments mainly reflect light from the red part of the
spectrum, perhaps accompanied by some from the orange or blue-
violet parts. A relatively large number of substances are known that
selectively reflect red light, but even so, our highest chroma red
pigments such as cadmium red reflect less light from the red part of the
spectrum than does a bright white pigment like titanium white (Fig.
4.5.5). Because chroma depends on the amount of light reflected, those
variants of cadmium red that reflect less light ("cadmium red deep"
etc) are reduced in chroma as well as value.

All high-chroma yellow materials reflect most of the red, orange,

yellow and green parts of the spectrum, and much more of their yellow
colour is due to additive mixture of the red and green wavelengths than
to the wavelengths that are yellow in themselves (Fig. 4.5.1). You will
see that such materials closely approach an optimal yellow colour, in
that they would pass on nearly all of the red and green spectral bands,
and absorb almost all of the blue-violet. Because wavelengths between
extreme spectral red and yellowish green mix additively along a
straight line in the CIE chromaticity diagram (Fig. 4.1.2), all mixtures
of these wavelengths remain at full saturation, and so bright yellow
materials, which reflect the whole of this range, are the lightest of all
high chroma materials. If less of the green part of the spectrum is
reflected, a lower-value and more reddish colour results, as in
cadmium yellow deep or cadmium orange. Conversely, if the
reflectance extends into the blue-green and cyan parts of the spectrum,
these wavelengths begin to neutralize the components at the red end,
producing a yellow that is greenish in hue and lighter-valued, but also
somewhat whitish and thus reduced in chroma (e.g. "lemon yellow").

Please note carefully that no bright yellow or orange substances reflect

or transmit mainly the yellow or orange wavelengths of the spectrum,
and absorb all other wavelengths. You may find quite a number of
websites and recent books on colour that get this wrong! The notion
that greenish-yellow and orange-yellow paints mainly reflect yellow
wavelengths, mixed with "impurities" or "traces" of green and orange
wavelengths respectively, is an old misunderstanding that has been
revived in recent years by a series of popular books that claim to reveal
"what really does happen when colors are mixed". No paints that
actually do this exist, and if they did they would reflect much less light
than a bright yellow paint, and so would be dark brown or olive in
appearance. It should really be clear that a bright yellow paint must
reflect most of the spectrum, because it is so close to white in value.
The green and orange wavelengths reflected by a yellow paint are not
impurities in the yellow; they are additive components of the yellow.

Figure 4.5.2. Reflectances of high-chroma green to blue paints at full strength

and as 2% tints with white.

While many different substances selectively absorb short-wavelength-

and reflect long-wavelength light, substances that do the reverse are
fewer in number and less effective in their action. None of
our green or blue paints attain the near-optimal plateau-shaped
reflectance curves seen in many red, orange and yellow paints. Our
strongest and most widely used green and blue pigments, the
phthalocyanines, reflect light of high saturation but very low brightness
at full strength, and higher brightness but compromised saturation
when mixed with white (Fig. 4.5.2). At an optimal level of white
content a maximum chroma of about 12 Munsell units is reached, but
this is moderate compared to the chroma of 14 to 16 of many red,
orange and yellow paints.
Figure 4.5.3. Reflectances of high-chroma violet-blue to magenta paints at full
strength and as 2% tints with white.

The difficulty in finding substances that selectively reflect short-

wavelength light also affects the range of available pigments in
the purple range. Magenta pigments such as quinacridone magenta
that have good reflectance in the red region have at best moderate
reflectance in the blue-violet region, and are therefore distinctly redder
than the equally blue and red digital magenta. More equal red-blue
reflectance is found only in substances reflecting much less red light.

Figure 4.5.4. Reflectances of moderate- to low-chroma earth pigment paints at

full strength and as 2% tints with white.
Figure 4.5.5. Reflectances of two white paints and two black paints, the latter at
full strength and as 2% tints with white.

Low-chroma substances generally have relatively flat reflectance

curves, in which all wavelengths are reflected to about the same level.
As with digital colours, the horizontal band spanning from the lowest
to the highest wavelength reflectance may be thought of as the coloured
"component" of the reflectance (for most actual substances, at least),
lying between a white component below and a black component above.
A large white component means that the object reflects light of low
saturation; a large black component means that the object reflects light
of low brightness; both these factors combine to lower the chroma.

It may be of interest to compare the spectral reflectance graphs of some

low-, medium- and high-chroma paints with the actual RGB values of
some images of those paints (Fig. 4.5.6). The RGB levels are shown
in linear units (proportional to light energy), obtained by raising the
standard nonlinear fractions to the power 2.2. For example a standard
(e.g Photoshop) R value of 128, or half of 255, becomes 0.5 to the
power 2.2, or 0.22. These linear RGB values in effect display a simple
spectral power distribution of the reflectance of the paint, in which the
spectrum is divided into just three bands.
Figure 4.5.6. Comparison of spectral reflectance curves of selected paints with
the actual linear RGB values of their colours on the screen.

Page added August 5, 2012.

<< 1 2 3 4 5>>

Next: Part 5: Subtractive Colour Mixing

PART 5. SUBTRACTIVE COLOUR MIXING

5.1 SUBTRACTIVE MIXING PROCESSES

 Subtractive mixing in coloured filters

 Subtractive mixing and coloured illumination
 Subtractive mixing in traditional paint media
 Subtractive mixing in digital painting

Subtractive mixing in coloured filters

As we have just seen, if we pass beams of light through two separate

coloured filters and then combine the beams, the resulting "colour
mixing" (strictly, colour stimulus mixing) is classed as additive
mixing. A different kind of "colour mixing" occurs if we instead overlap
the two filters and pass a single beam through them together. Now the
effect of the second filter is to remove light that has passed through
the first filter, rather than to add to that light, and so the process is
generally known as subtractive colour mixing. The resulting beam
consists of those wavelengths that are transmitted by both filters, as
opposed to either filter in the case of additive mixing (Fig. 5.1.1).
Figure 5.1.1. Additive and subtractive mixing contrasted using ideal yellow and
cyan filters. Adding cyan and yellow light gives greenish white
light; successively removing light with cyan and yellow filters gives saturated
green light.

The result of subtractive mixing of two filters is calculated by

multiplying together the percentage of light energy passed on by both
filters, for each wavelength of the visible spectrum. (This has led some
to prefer the term multiplicative mixing). Thus if 20% of a particular
wavelength is passed on by one filter and 50% of the same wavelength
is passed on by the other, the subtractive result will be 20% of 50%, or
10%. If any wavelength is completely blocked by one or other filter,
then that wavelength will be absent from the result. In the theoretical
example of Fig 5.1.1, an ideal yellow filter transmits all light in the red-
orange and green bands of the spectrum, but no blue-violet light, while
the ideal cyan filter transmits all light in the green and blue-violet
bands, but no red-orange light, and so subtractive mixing leaves only
the spectral component that the two filters have in common: green.

Take a moment at this point to make sure that you have completely
eradicated from your mind the naive idea that green is "made of"
yellow and blue. Subtractive mixing doesn't work like that. We get a
green mixture from yellow and cyan because these components both
are partly (so to speak) "made of" green. If any colour can be said to be
"made of" yellow and blue, it's white!
Subtractive mixing and coloured illumination

The colour stimulus resulting from the interaction of a coloured

object with a coloured light source is another example of subtractive
mixing (Fig. 5.1.2). Here we think of the incident light as having had
certain wavelengths removed compared to white light, and we multiply
the percentage of each wavelength present with the percentage of that
wavelength reflected by the object. How this stimulus is
actually perceived, however, will depend on the degree to which the
observer's vision adapts to the colour of the light source.

Figure 5.1.2. Cyan coloured ball under yellow light. IMAGE: D. Briggs,
Photoshop CS2.

Subtractive mixing in traditional paint media

Subtractive mixing is important to artists using traditional paint media

because of its role in the physical mixing and glazing of paints.
Subtractive mixing is involved in mixing or glazing of physical
colourants because most light will be influenced successively by
particles of each component colourant (Figs 6.1.2, 6.1.3). Once again
the result of the subtractive interaction is calculated by multiplying the
percentages of light passed on by each colourant, wavelength by
wavelength throughout the spectrum.
Because of the phenomenon known as metamerism, two colourants
may be similar in colour, but have somewhat different spectral
reflectance curves, and consequently the exact results of subtractive
mixing of real colourants can not be predicted merely from their
colour. This uncertainty should not be overstated, however. All
common cyan and yellow colourants combined subtractively will make
a green; only the precise appearance and spectral composition of that
green will depend on the specific colourants used.

Subtractive mixing in digital painting

Subtractive mixing in digital media is emulated by blending layers

in Multiply mode, which multiplies the RGB values of a layer with the
underlying RGB values. This blending operates on the nonlinear
(perceptual) RGB values, but because it is a multiplication, the
resulting colour is the same as if it were done on the linear (physical)
RGB values. Subtractive mixing is simpler in digital media than in
traditional media and in nature, in that there are only three bands of
the spectrum to be considered. It gives a realistic representation of
what subtractive mixing involving comparably coloured lights and
materials might result in, but real-world subtractive mixing could give
a somewhat different result because it depends on the actual spectra of
the lights and materials involved. Unrealistic effects may result from
subtractively mixing very bright and/or very saturated digital colours
that are outside the range of real object colours.

Even in a program such as Painter, which cleverly simulates the

appearance and physical behaviour of artists' paints, colours
nevertheless still mix by ideal subtractive rules (e.g. "Monitor yellow"
and "Monitor blue" mix to make black or grey, while paints of similar
hues would mix to a dull green).

Modified August 13, 2012. For original (2007) page see here.

<< 1 2 3 >>
5.2 IDEAL SUBTRACTIVE PRIMARIES

Imagine a set of three hypothetical, perfect filters, each transparent to

about a third of the spectrum, and completely opaque to the
remainder, constructed such that they each let through a band of
wavelengths like one of the red, green and blue optimal colour additive
primaries in Fig. 4.1.1. These perfect filters would be extremely difficult
to manufacture from physical colourants, but in the digital realm our
full-chroma red, green and blue colours work exactly like this when
they are mixed subtractively (in Multiply mode). A moment's thought
will show that these ideal red, green and blue filters would be
completely unsuitable as primaries for subtractive mixing. These filters
by definition have no wavelengths in common, and so any pair at full
strength would absorb all light, making black, and changing the
strength of one of the filters would not result in intermediate hues (Fig.
5.2.1).

Figure 5.2.1.
Subtractive mixing of ideal colourants corresponding to the optimal
colour additiveprimaries. Filters that let through only one additive primary at
a time would not work for subtractive mixing. Varying the strength of the red
(top slider) , green (middle slider) and blue (bottom slider) colourants does not
produce intermediate hues. Rectangles show the phosphors glowing in each
filter and (top left) in the overlap of all three. Copyright David Briggs and Ray
Kristanto, 2007.
To work as perfect primaries for subtractive mixing, the filters would
instead need to completely transmit two of the additive primaries at a
time, and completely absorb the third. There are just three pairings
possible, R+G, R+B, G+B. As we saw in Section 4, these pairings are
seen as yellow, magenta and cyan respectively. Each of the three
possible combinations of these pairings has one and only one of the
additive primaries in common. Yellow (R+G) and magenta (R+B), for
example, having red in common, and at full intensity mix subtractively
to make red (Fig. 5.2.2). Changing the intensity of one or other of the
components produces a continuous range of hues between yellow and
magenta (Figure 5.2.3), and the corresponding yellow to cyan and cyan
to magenta intermediates complete the circuit of possible hues. No
wavelength is shared by all three primaries, and so subtractive mixing
of the three at full intensity produces perfect black.

Figure 5.2.2. Mixing of ideal subtractive primaries at full intensity.

 Subtractive mixing of ideal subtractive primaries. Varying the strength
Figure 5.2.3.
of the cyan (top slider) , magenta (middle slider) and yellow (bottom slider)
produces a full range of intermediate hues. Rectangles show the phosphors
glowing in each filter and (top left) in the overlap of all three. Copyright David
Briggs and Ray Kristanto, 2007.

In the digital realm pairs our subtractive primaries can mix to make all
possible RGB colours, and so these mixing relationships can be
schematically expressed as a triangle (Fig 5.2.4).
Figure 5.2.4. Mixing relationships of ideal subtractive primaries.

Note however that in Munsell colour space these colours occupy a

different and less regular shape, with maximum chroma at
the additive primaries (Fig. 5.2.5, black outline).
Figure 5.2.5.
Comparison of mixing paths of "ideal" primaries and gamut of sRGB
colours (i.e. of sRGB subtractive primaries) with chroma limits of Munsell Book of
Color (black dots) and limits of theoretically possible colours (Macadm limits),
plotted using Zsolt Kovac's program drop2color.

In simplistic colour theory we are often informed that "in theory" we

should be able to mix all colours from paints of the three primary
colours, but that in practice this is not possible because our paint
primaries are "imperfect". The implicit "theory" here is what I refer to
as the naive conception of the primary colours: that colours are
physical properties residing in objects, and that all colours are
physically "made of" red, yellow and blue. In reality it is not possible
for just three colours to subtractively mix all possible colours, even
theoretically. To illustrate the mixing of ideal subtractive primaries in
his color2drop programme, Zsolt Kovacs devised three ideal primaries
with slightly overlapping spectra, but close in general hue to the RGB
subtractive primaries. These ideal primaries mix a much larger gamut
than is obtained by physical mixing of actual yellow, magenta and cyan
paints, but still do not mix even the entire range of common object
colours, as indicated by the limits of the Munsell Book of Color, let
alone the entire range of theoretically possible colours. The same is
true of the gamut of our digital subtractive primaries, for example the
sRGB primaries also shown in Fig 5.2.5.

Modified August 12, 2012. For original (2007) page see here.

<< 1 2 3 >>

5.3 SUBTRACTIVE COMPLEMENTARIES

 Subtractive vs additive complementaries

 Subtractive mixing paths

Subtractive vs additive complementaries

We've seen that the term "complementary" in additive mixing refers to

colours of pairs of lights that mix to make white light. This white light
may be seen as white light, or in the case of optical mixing, as a white
or a grey object colour. In subtractive mixing, "complementary" also
refers to pairs of colours that produce a neutral mixture, which is
usually seen as a grey or a black object colour.

The subtractive complementary of a colour is frequently not the same

as its additive complementary, even in the relatively simple case of
subtractive mixing of digital colours. Consider an RGB colour with
linear (i.e. energy-proportional) RGB of a b c, where a,b, and c are
expressed as fractions of the maximum R, G and B respectively.
Additive complementaries of abc would include the RGB colour 1-a 1-b
1-c, plus all colours differing from this only in brightness and/or
saturation (that is, all colours of the same hue angle). Subtractive
complementaries of abc however would include all colours with R,G
and B in the ratio 1/a: 1/b:1/c, because these would multiply
with abc to give a neutral result; this set of colours would have a
common hue angle and saturation, but would vary in brightness.

As an example (Fig. 5.3.1), consider a purple with linear RGB of 0.60

0.20 1.00 (or 153 051 255), and a green with linear RGB of 0.40 0.80
0.00 (or 102 204 000). These two colours are additive
complementaries because they can combine (in the right proportions)
to make white light, as would be any other pair of RGB colours of the
same two hue angles. They are not subtractive complementaries
however, because they would not multiply to bring all three
components of the purple colour to the same level. A green that would
do this is one with linear RGB of 0.33, 1.00, 0.20 (or 085 255 051).

Figure 5.3.1. Additive and subtractive complementaries of a single colour

differing in hue angle by 30°. The additive complementary adds to make white
light; the subtractive complementary multiplies the R,G and B brightnesses to
make grey.

Subtractive mixing paths

Subtractive mixing paths are more complicated than additive mixing

paths, but in the relatively simple situation of mixing of "full-chroma"
digital colours they display a distinctive pattern that turns out to
persist clearly even in the much more complex situation of physical
paint mixing. In the interactive demonstration Fig. 5.2.1, we saw that
"Monitor Red" (R 255) was subtractively neutralized to black both by
'Monitor Green" (G 255) and by "Monitor Blue" (B 255). Thus by the
definition given at the top of this page, both of these colours are
subtractive complementaries of "Monitor Red", as are all intermediate
full-chroma colours passing through "Monitor Cyan". For the additive
primaries "Monitor Red", "Monitor Green" and "Monitor Blue", all full-
chroma colours between 120o and 240o away on the digital hue circle
completely lack that primary, and so, having no component in
common, are its subtractive complementaries. Colours up to the
nearest subtractive primary, 60o away on either side, mix at full
chroma (Fig. 5.3.2).

Figure 5.3.2. Subtractive mixing of a full-chroma additive primary with other

"full-chroma" digital colours. Mixing paths displayed in plan view of YCbCr
space, using the program ColorSpace by Philippe Colantoni.

Conversely, all full-chroma colours between two additive primaries

have a single complementary, the third primary at full chroma, and
therefore full-chroma subtractive primaries like "Monitor Magenta",
which mix at full chroma with colours up to 120oaway on either side,
have only one complementary, directly opposite in hue . The resulting
pattern of mixing paths is very distinct from that of a full-chroma
additive primary like "Monitor Red" (Fig. 5.3.3).
Figure 5.3.3. Subtractive mixing of a full-chroma subtractive primary with other
"full-chroma" digital colours. Mixing paths displayed in plan view of YCbCr
space, using the program ColorSpace by Philippe Colantoni.

Despite all of the additional complications involved, these two patterns

seen in the simplest kind of subtractive mixing are still recognizable in
the mixing paths of actual artist's paints (Fig. 5.3.4), as we will see in
the next section.
Figure 5.3.4. Comparison of ideal subtractive mixing patterns of full-chroma
digital colours with mixing paths modelled for actual paints in the Munsell hue-
chroma plane using the program drop2color by Zsolt Kovacs.

Modified June 26, 2012. For original (2007) page see here.

<< 1 2 3 >>

Next: Part 6: Colour Mixing in Paints

PART 6. COLOUR MIXING IN PAINTS
6.1 MIXING PAINTS

 Paint mixing processes

 Paint-mixing primaries

Paint mixing processes

Artists who use physical paint media mix their colours in three basic
ways:

1. by physical mixing,
2. by glazing (using superimposed transparent paint layers), and
3. by interspersing small patches of colour that optically blend
completely or partially at the intended viewing distance.

The physical mixing of paints is often described simply

as subtractive, but in reality a component of additive-averaging mixing
is usually also involved. In a physical mixture of paints, most light
bounces around and interacts with grains of all components in the
mixture before emerging, so that each component has the opportunity
to influence the colour of the light. However, unless the paints involved
are perfectly transparent, some light will be back-scattered off grains of
one or other component alone (Fig. 6.1.1). The first process is
subtractive, whereas the second by itself would result in additive-
averaging mixing. Thus even if two opaque paints reflect no
wavelenghths in common, their mixture will still reflect some light
because of the contribution of the additive-averaging component,
which is why it is impossible to mix a deep black from them.
Figure 6.1.1. Physical mixing of artists' paints. Physical mixing of opaque paints
involves a substantial component of additive-averaging as well as subtractive
mixing. With transparent pigments the process is more purely subtractive.

Figure 6.1.2A compares the physical mixing path of two near-

complementary paints with the paths of additive, additive-averaging,
and subtractive mixing of their RGB colours. (Remember from Section
4.5 that the linear RGB values of a paint's colour constitute a simple
graph of the spectrum of light that the paint reflects). For the opaque
paints used in this illustration, the paint-mixing path is intermediate
bewteen the subtractive and the additive-averaging path. For more
transparent paints, the path would be closer to the subtractive path.
When grains of the same pigments are not enclosed in a medium, as for
example when colours are blended in pastel, the mixing path is visually
closer to the additive-averaging path (Fig. 6.1.2B). This is because an
enclosing medium makes the pigment grains more transparent, and
also, due to internal reflection within the medium, increases the
amount of multiple reflection among the grains, two factors that favour
the subtractive component.

Figure 6.1.2. Mixing paths of two near-complementary opaque pigments,

cadmium red and cobalt green, (A) ground in linseed oil, and (B) mixed as dry
powders, compared with the paths of their RGB colours mixed on additive,
additive-averaging and subtractive principles. The latter were calculated in
Photoshop using opacity gradients in Linear Dodge, Normal and Multiply layer
mixing modes respectively, blended at a gamma of 1.0. Notice that the pigment-
mixing paths are a compromise between subtractive and additive-averaging
mixing, visually closest to the subtractive path in paint mixing, but to additive-
averaging path in powder mixing.
In glazing (Fig. 6.1.3), most light will pass through both layers,
resulting in subtractive mixing. However some light will be scattered
by the top layer unless the latter is perfectly transparent, and because
of this a red glaze over blue looks different to a blue glaze over red.

Figure 6.1.3. Glazing of imperfectly transparent paints. Mixing of overlaid

glazes is primarily subtractive, but a minor additive-averaging component is
supplied by the upper layer.

Use of minute interspersed colour areas (Fig. 6.1.4) results in

primarily additive-averaging colour mixing. Thus, even though physical
paints are involved, an area of yellow and ultramarine blue dots in the
right proportion will make a neutral grey, rather than the green that
would result from subtractive mixing. However if there is any
overlapping of transparent component colourants (as occurs
extensively in halftone printing), or mutual reflection between the
coloured areas, subtractive mixing will also be involved.

The pointillist painting technique of applying separate dots of colour

that partially fuse optically at the intended viewing distance was
advocated by Ogden Rood (1879) as a means of producing a "soft" but
vibrant visual effect. There are rather widespread notions that in the
work of Neoimpressionist Georges Seurat this technique produces, or
was intended to produce, brighter and/or more intense colours than
could be obtained from "palette mixtures". This is a misunderstanding
of both Rood and Seurat. Seurat in fact employed optical colour mixing
most elaborately in the most neutral areas of his paintings, where it
allowed the painter to produce low-chroma colour fields while still
employing high-chroma visual components. Related effects of colour
"noise" are also be obtained by letting colour show through
discontinuities in an overlying paint layer, or in watercolour by using
washes containing a mix of granulating and staining components.
Figure 6.1.4. Mixing of minute interspersed colour areas. This kind of mixing is
mainly additive-averaging, with a subtractive component if the components
interact by overlap or inter-reflection.

Paint-mixing primaries

We saw in Section 5.3 that in the relatively simple subtractive mixing

seen in digital colours, the full-chroma additive and subtractive
primaries each mix with other full-chroma hues in a distinctive pattern
of mixing paths (Fig. 6.1.5).

Figure 6.1.5. Subtractive mixing patterns of (A) full-chroma subtractive

primaries and (B) full-chroma additive primaries in a plan view of YCbCr space,
which provides a regular hexagonal representation of the hue-chroma plane of
the RGB gamut (see section 5.3).

Mixing of artists' paints is more complex, both because the subtractive

result depends on the actual spectra of the pigments involved, and
because of the additional additive-averaging component discussed
above. Despite these "substance uncertainties", the two basic patterns
delineated in the simplest kind of subtractive mixing are still
prominent in the mixing paths of artist's paints (Figs 6.1.6, 6.1.7,
6.3.1A-G). Paints that are distant from the ideal subtractive yellow,
magenta and cyan hues, for example cadmium red, make high-chroma
mixtures with only a few close hues, and have near-complementaries
among a surprisingly wide range of hues, such that several paints may
serve as paint-mixing complements for them. Mixing paths funnel
towards these colours in a pattern that I liken to an introverted
octopus (Fig. 6.1.6). In contrast, paints that are close to ideal yellow,
magenta or cyan hues follow mixing paths shaped like an extroverted
octopus: they make high-chroma mixtures with distant hues, and are
directly neutralized by only a very narrow range of paint hues (Fig.
6.1.7). If an exact mixing complementary is needed for one of these
subtractive primaries, it will often need to be mixed, as there is a
comparatively low probability that the precise hue required will be
found at hand in a given paint selection..
Figure 6.1.6. Mixing paths of artist's paints close to the additive primary hues.
(A) Mixing paths of Liquitex Acrylic Cadmium Red Light with other paints in
the range, modelled using the program drop2color by Zsolt Kovacs. (B) Mixing
paths of Art Spectrum Cadmium Red oil paint with other paints of various
brands; all paints raised to approximate maximum chroma point with added
white. RGB image colours plotted in CIE L*a*b* space using the
program ColorSpace by Philippe Colantoni.
Figure 6.1.7. Mixing paths of artist's paints close to ideal subtractive primary
hues. (A) Mixing paths of Liquitex Acrylic Medium Magenta with other paints
in the range, modelled using the program drop2color by Zsolt Kovacs. (B)
Mixing paths of Art Spectrum Quinacridone Magenta oil paint with selected
other paints of various brands; all paints raised to approximate maximum
chroma point with added white. RGB image colours plotted in CIE L*a*b* space
using the program ColorSpace by Philippe Colantoni.

Because paints close to the ideal subtractive primary hues mix along
lines that bend outwards, staying high in chroma, any set of three such
paints yields a particularly large gamut or range of colour mixtures
(Fig. 6.1.8). Yellow, magenta and cyan are thus the optimal primary
hues for paint mixing, just as they are for standard colour printing and
photographic prints and slides. The growing realization of the
particular effectiveness in colour mixing of three such colours among
painters and dyers of the early sixteenth century inspired the concept
of primary colours, although historically these primaries were
generally identified by the names yellow, red and blue (Section 6.2).

Ideal yellow, magenta and cyan colourants would be the optimal

primaries for paint mixing, if such pigments existed. However, even
today our best magenta and cyan pigments are well towards red and
blue respectively, compared to the ideal magenta and cyan subtractive
primaries. No high-chroma pigments are very close to an ideal magenta
hue; the closest high-chroma oil painting pigment is quinacridone
magenta (PV 19), which is distinctly redder than ideal magenta. The
closest popular oil painting pigment to an ideal cyan is the green shade
variant of phthalocyanine blue (PB 15.3), which similarly is distinctly
bluer than an ideal cyan. Cobalt green is closer to ideal cyan in hue,
and does mix some blue-green colours not obtained using
phthalocyanine blue alone as the primary, but yields fewer colours in
the bluish range. It is also more expensive, and being opaque is less
suited to mixing dark colours. Yellow pigments are available in an
essentially continuous range of hues, and a pale or lemon yellow seems
to provide the optimal gamut of colours.

Figure 6.1.8. Mixtures of (A) Lemon Yellow (B) Quinacridone Magenta and (C)
Pthalocyanine Blue (Green Shade) oil paints. Photographed colours in a*b*
plane of L*a*b*space. Each pair of primary colours mixes along a line that is
convex outwards, keeping relatively close to full chroma.
Figure 6.1.9. Mixtures of (A) Cadmium Orange, (B) Ultramarine Blue and (C)
Pthalocyanine Green (Yellow Shade) oil paints. Photographed colours in a*b*
plane of L*a*b*space, viewed from below. Colours on each mixture line move
away from full chroma.

Modified 2nd November, 2012. Original text here and here.

<< 1 2 3 >>

6.2 PRIMARY COLOURS

 Theories of primary colours

 The historical primaries: yellow, red and blue
 "Split-primary" palettes

Theories of primary colours

The expression "primary colour" has its origin in the historical concept
that yellow, red and blue, initially alongside white and black, were the
"simple", "primitive" or "primary" colours from which all others could
be derived by mixing. The idea that painters can mix all colours except
three can be traced back to Aristotle in his Meteorologica [c. 350 B.C.],
but surprisingly Aristotle (Fig. 6.2.1A) gives these colours as the same
three he saw in the rainbow: red (phoinikoun), green (prasinon) and
blue/violet (alourgon). Yellow, red and blue are placed between white
and black in a linear scale mentioned in a commentary on
the Timaeus of Plato from the fourth/fifth century CE (Kuehni, 2003),
and the same scale also appears in the first visual representation of the
concept of primary colours, a diagram in Francois
D'Aguilon's Opticorum Libri Sex of 1613 (Fig. 6.2.1B).

Figure 6.2.1. A. Extract from the Meteorologica of Aristotle (tr. H. Lee, 1952,
Loeb Classical Library). B. From Francois D'Aguilon, Opticorum libri sex of
1613. The "simple" colours albus, flavus, rubeus, caeruleus and niger (white,
yellow, red, blue and black) are placed on a linear scale, and aureus,
viridis and purpureus (gold, green and purple) are generated from mixtures of
the middle three. Picture credit: Institute an Museum of the History of
Science, Biblioteca Digitale

The yellow-red-blue primaries seem to have rapidly gained popularity

in the practice of artists and dyers in the first decades of the
seventeenth century, and are increasingly recorded in print over the
course of the century (Shapiro, 1994; Kuehni, 2010). Robert Boyle
(1664) introduced the term "primary colour" in English for these
colours in a statement that shows an awareness of the concept of
a gamut: while the primary colours suffice to mix colours of a full
range of hues, some colours will, by their greater "splendor" (we would
say chroma), lie outside this gamut (Fig. 6.2.2).
Figure 6.2.2. From Experiments and Considerations Touching Colours (1664)
by Robert Boyle.

It is a small and slippery step from the observation that all hues can be
made from three primary colours, to the assumption that all hues are
made of those three colours. This latter belief, which I call here
the intermixture model of primary colours, rests on what some
philosophers call the "commonsense" view of colour, that colours are
physical properties residing in objects, combined with the further
"commonsense" assumption that these physical properties themselves
intermix when colourants are mixed, i.e., that "colour mixing" is
literally mixing colours. On these assumptions, the observation that
green paint results from mixing yellow and blue paint leads to the
interpretation that green is a secondary/ compound/ mixed
colour that is "made of" yellow and blue. If bright yellow paint can not
be made by mixture, then that is because yellow is a primary/ simple/
pure colour that can not be compounded from other colours. If a
secondary colour and the remaining primary together "contain" all
three primaries, then so must any mixture made from them. Grey,
black, and all greyed colours are therefore held to contain all three
primaries (hence the term "tertiary" colour in its original sense). If all
colours are compounds of the three primaries, then "in theory" we
ought to be able to mix all colours at full chroma from these primaries.
The fact that we can not is interpreted to mean that our paint primaries
must be imperfect, or biased with impurities of other primaries; when
mixed these impurities supposedly combine to constitute a black or
grey component that dulls the mixture.
Figure 6.2.3. Extracts from Johannes Itten's The Art of Color (1973, pp. 34, 78
and 22). Although Itten acknowledges Newton and the spectral decomposition
of light in an introductory page, this does not disturb the intermixture model
that prevails throughout the rest of the book, where green is a "mixed color" that
is "composed of" or "contains" yellow and blue. Later in the book (p. 137) he
maintains that the expressive meanings he attaches to colours compound in the
same way: e.g. compassion (green) = knowledge (yellow) + faith (blue).

It is not always possible to be certain what a historical writer actually

believed about colour mixing, but the intermixture model is likely to
have been widely assumed by the early proponents of terms like
"primary"/ "primitive" and "secondary", "simple" and "compound",
and "pure" and "mixed" colours. In 1613 D'Aguilon had distinguished
three types of intermixture, "real" (compositio realis), for the mixing of
materials such as paints, "intentional" (compositio intentionalis), for
mixing involving a transparent medium, such as glazing, and
"notional" (compositio notionalis) for what we would call optical
mixing (Kemp, 1990, p. 276). In 1708, the anonymous author of
the first hue circle of artists' paints divided colours into primitive
colours ("Couleurs Primitives") "that cannot be composed of other
colors", mixed colors that are mixed from these primitive colors, and
other colours "that are composites of the primitive colors and the
mixed colors" (translated Kuehni, 2010). The intermixture model of
primary colours is explicit in the works of nineteenth century theorists
such as Goethe and George Field, and is alive and well in art and design
circles today. In many minds it exists in quarantine alongside
knowledge of Isaac Newton and the spectrum, without any apparent
awareness of the inconsistency, as it does in Johannes Itten's
(1961) The Art of Color (Fig. 6.2.3).

Newton proved that the intermixture model of primary colours was

untenable as a physical explanation by showing that light of all hues
except extraspectral purple could be either simple
(monochromatic) or compound. Further, he showed that compound
light is always compounded from the innumerable but five- or seven-
named rays of the spectrum, and that objects of all hues reflect
compound light containing more of some of these spectral rays than
others. Although Newton explicitly stated that colour itself is a mental
perception, he pictured these physically simple and compound lights as
quite directly evoking simple and compound colour sensations in the
observer via "vibrations" in the optic nerve. He therefore regarded the
colours of the spectral rays as the fundamental components of all
colour experience, and hence considered these to be the true primary
colours (Fig. 6.2.4A).

Figure 6.2.4. Two extracts from Newton's New Theory (1672).

Though often criticized for failing to distinguish between additive and

subtractive mixing, Newton in fact gave the first correct explanations of
subtractive processes, for example explaining that a blue powder under
a "compound" yellow light (Fig. 6.2.4B) looks green because green rays
are the element in common between the light and the reflectance of the
powder, and thus not because the green is "made of" yellow and blue.
(D'Aguilon had specifically given the green appearance of a blue
powder by candle light as an example of compositio intentionalis).
However in describing the mixing of coloured powders Newton
accounted for the darkness of the neutral mixtures he obtained with an
explanation essentially in terms of additive-averaging (each grain
absorbs some light, even of its own colour, whereas in a white powder
all grains reflect all of the light). Additive-averaging actually is the
dominant process involved in the mixing of powders (Fig. 6.1.2), but a
secondary role is played by subtractive mixing, which Newton did not
invoke here, and Newton said almost nothing about paint mixing,
where the subtractive process is dominant. Newton's theory of
continuous spectral primaries therefore left a major problem
untouched: if the spectrum was continuous, how could we explain the
three primary colours of the painters and dyers? Why
should three particular colours in that continuum be optimal for
mixing colourants?

One plausible solution was that Newton was wrong, and that the
apparently continuous spectrum was actually made up of red, yellow
and blue rays that overlapped to produce the intermediate hues by
intermixture. This theory of three spectral
primaries incorporated Newton's discoveries of the spectral
decomposition of white light and the role of selective absorption in
causing object colours, while providing a simple rationalization for the
three colourant primaries that left the intermixture model intact. For
example, a paint mixture containing yellow and blue particles could be
assumed to reflect yellow and blue rays that intermix to make green,
just as they were thought to do in the spectrum. This theory was held
by many 18th and 19th century innovators in the field of colour,
including Mikhail Lomonosov and George Palmer (both of whom
proposed the existence of three types of colour vision receptors based
on the assumption of three spectral rays), astronomer Tobias Mayer
(who invented the first fully described colour space based on the same
assumption), and Ducos du Hauron (inventor of the first subtractive
colour photographs). The discoverer of the additive primaries of light,
Christian Ernst Wunsch, proposed a second theory of three spectral
primaries, in which the spectrum was instead held to be made of
overlapping red, green and violet rays (Wunsch, 1792).
Figure 6.2.5: A. Colour space in the form of a double pyramid devised by
Swedish astronomer Tobias Mayer, alongside Mayer's diagram of the triangular
red-yellow-blue basal plane, and a hand-coloured copy of one of Mayer's planes
by Lichtenstein. B. George Palmer's theory of three visual receptors, based on
assumption of three "rays of light" (Palmer, 1777).

In his monumental Zur Farbenlehre (1810), the German poet Johann

Wolfgang von Goethe ridiculed the spectral primary theories of both
Newton and Wunsch. For the origin of colours he proposed a theory
reviving the idea that colours are formed by modification of white light
at boundaries between light and darkness. His theory invoked two
elementary primaries, yellow and blue, both of which "augment"
or intensify in the direction of pure red. When "arrested" in the
physical state, yellow, red and blue form the painter's three
"primitive" colours (ibid., 705). Unlike Newton, Goethe considered
colours, excluding "physiological" and related colours, to be physical
properties residing in ("arrested in") objects and light, and held that
these colours themselves combine by intermixture. Goethe included
processes now regarded as additive, subtractive and perceptual in a
category he called "real intermixture", while classifying optical mixing
and tentatively pigmentary mixing as "apparent intermixture" (ibid.,
551-571). Finally, in his discussion of the nomenclature and "moral
associations" of colours, Goethe devised a framework based on what
might be called his four nomenclatural primaries, yellow, blue,
red and green (ibid., 610-611).

Figure 6.2.6. A. A beam of light emerging from a prism exhibits a yellowish and
a bluish coloured fringe separated by a short wedge of white light. Goethe (1810,
pl. 5) maintained that the two coloured fringes were created at the boundaries of
light and darkness, and subsequently mixed in the middle of the spectrum to
make green, while each "augmented" outwards towards red. B. Newton had
already explained in the Opticks (Book I, Part II, Fig. 12) that this white wedge
was formed by additive mixing of the overlapping spectra formed by successive
rays within the beam. If Goethe was right, the coloured rays should continue
crossing over further from the prism, while if Newton was right the rays should
become more and more cleanly separated. Newton's experiments used spectra
projected over distances that were very long (up to 22 feet) compared to the
length of the white wedge, and thus disproved Goethe's theory of the origin of
colours before Goethe was born.

The theory of three spectral primaries reached its peak of influence in

the early 19th century through the efforts of the formidable Scottish
physicist David Brewster (1781-1868), who convinced himself he could
see the red, yellow and blue components of light by viewing the solar
spectrum through coloured filters (Fig. 6.2.7). Brewster's dogmatic
advocacy of his idea made the theory of three spectral primaries into
something of a roadblock that needed to be demolished before progress
in colour research could proceed (Sherman, 1981). This demolition was
accomplished, elegantly and finally, by Helmholtz (1852b), who
accounted for Brewster's observations in terms of a combination of
imperfect experimental conditions and visual contrast effects. The
theory of red, yellow and blue spectral primaries nevertheless
remained current in artistic circles well into the twentieth century, and
strongly influenced the thinking of several early abstractionists. As late
as 1975 Josef Albers still thought it was regarded as a "plausible theory"
(Albers, 1975, p. 23).

Figure 6.2.7. A. Brewster's (1831) interpretation of the solar spectrum in terms

of overlapping red, yellow and blue rays. B. Simulation of Brewster's purported
observation.

Fifty years before Helmholtz's demolition of Brewster's theory, a

solution to the problem posed by the three-fold nature of the primaries
had been offered by Thomas Young, who suggested that it reflects the
structure of human colour vision and not the physical structure of light.
Young's held that Newton was correct about the continuous spectrum,
and that the number of primaries might be giving us a clue to the
number of types of colour vision receptors (now called cone cells) in
the eye, which after all would surely have to be finite. At first (Young,
1802) he suggested that the three types of receptors were tuned to the
"principal colours" red, yellow and blue, but within months he
switched to the spectral additive primaries, red, green and violet. As he
later explained (Young, 1807, p. 439), since it was known that yellow
light could be mixed from red and green light, and blue from green and
violet, the simplest solution was that red, green and violet were the
fundamental sensations from which all other colour sensations were
compounded (Fig 6.2.8A). Helmholtz (1852a) initially questioned this
and held that there would actually need to be at least five simple
colours, but later (Helmholtz, 1866) realized that there could be just
three receptors if these had overlapping sensitivities. In his 1866
visualization of the theory (Fig 6.2.8B) the receptors responsible for
the red and violet "sensations" have their peak sensitivities in parts of
the spectrum that are seen as red-orange and violet-blue respectively
(due to admixed green "sensation").

Figure 6.2.8. A. Thomas Young's (1807) diagram of the spectrum. B.

Helmholtz's (1866) depiction of Young's theory.

The Young-Helmholtz theory of three fundamental

sensations was supported and greatly refined in experimental colour-
matching studies by James Clerk Maxwell and Arthur Konig among
others, which led utimately to the development of modern
colourimetry. The modern CIE system defines three specific additive
primaries, the CIE 1931 Standard Primaries, consisting of narrowband
light sources of wavelength 700 nm (red), 546.1 nm (green) and 435.8
nm (violet). These are converted mathematically to three purely
theoretical lights, X, Y and Z, that allow all colour stimuli to be
matched mathematically without needing to use negative values; these
CIEÂ "primaries" do not correspond to actual colours or components of
colours. Both the monochromatic standard primaries and the
imaginary X,Y and Z "primaries" are convenient but ultimately
arbitrary choices.

Young and Helmholtz, like Newton, pictured a quite direct relationship

between visual stimulus and colour sensation, and for adherents of the
Young-Helmholtz theory, red, green and violet or blue replaced the
"Brewster primaries" of yellow, red and blue as the "true" primary
colours, which created all other colour experiences by intermixture.
Even today, either the additive primaries or the three cone responses
are still often assumed to be the "real" primary colors. However, as we
have seen in the section on colour vision, the Young-Helmholtz theory
is now combined with Ewald Hering's opponent theory as the zone
theory, which holds that our colour vision involves trichromatic
input from three cone cell types, and opponent output consisting
of yellow/blue and red/green opponent signals. According to the zone
theory, the three cone types do not "detect" red green and blue, but
instead each responds to most or all of the spectrum to varying
degrees, and the brain creates red/green and yellow/blue colour
perceptions based indirectly on differences in cone responses.
Hering's four opponent hues (Fig. 6.2.9) are often called the four
psychological primaries because they are the only hue perceptions
that can exist in a "unique" (pure) state, all other hues being mental
mixtures of adjacent pairs of these hues. They are thus the true
primary colours in the most literal sense. This set of four hues was
anticpated in the much earlier systems of Leonardo and Forsius, and
both the hues and their oppositions were anticipated in Goethe's
concepts of two elementary and four nomenclatural primaries (see
above).

Figure 6.2.9. Ewald Hering's analysis of the hue circle into red/green and
yellow/blue components. His four primaries of colour experience are now
reconciled with Helmholtz's three primaries of colour stimulus in the zone
theory of colour vision.

Helmholtz's (1852a) explanation of the role of subtractive processes in

paint mixing, and the recognition of the inverse relationship between
additive and subtractive primaries, allowed the definition of
three optimal subtractive primaries, under such names
as yellow, pink and seagreen (Benson, 1868), yellow,
magenta and cyan blue (Hatt, 1908), or Zanth, Achlor, and Syan (Ives,
1935). Yellow, magenta and cyan, which are still sometimes known as
the Ives primaries, have long been accepted as the optimal primary
hues for three-colour printing and photographic prints and slides,
although in these fields they often remained known up to the mid-20th
century by the names of the historical primaries, yellow, red and
blue. The latter are still taught as the primary colours in many art and
design classes today.

Figure 6.2.10. A. Inverse relationship of additive and subtractive primaries,

from Benson (1868). B. Hue circle based on yellow, magenta and cyan blue
primaries, from Hatt (1908).

The historical primaries: yellow, red and blue

We've seen that the expression "primary colour" is currently applied in

several contexts1beyond the original one of colourant mixing, and that
these primaries are different in each case:

 Y, B, R, G: the four psychological primaries. These result from

the opponent output of our visual system as yellow/blue and
red/green signals. We experience these and only these as pure
hues, and all other hues as mixtures of these.
  "R" (red/ orange-red), "G" (green/ yellow-green), "B" (blue/
blue-violet/ violet): the three additive primaries, i.e. hues of
lights giving a large gamut of additive mixtures. These primaries
reflect the trichromatic input into our visual system: "R", "G" and
"B" lights each stimulate one of the three cone types more than
the other two.
 Y, BR (ideal magenta), GB (ideal cyan): the
three theoretical optimal subtractive primaries. These have a
direct, complementary relationship to the additive primaries.
 Y, bluish-R (process magenta*), greenish-B (process cyan*):
the three best primary colourant hues actually available. These
primary hues apply equally to inks and to paints, though
different substances may or may not be involved. (*Before the
20th century, Prussian blue and insect or madder crimson
respectively).

The historical primaries, Y, R and B are the three names that we

generally applied to our best primary colourant hues when we first
discovered them, before it occurred to us that the hues we mentally
experience as pure and primary might not be the same as the optimal
primary hues for colourant mixing. (And this is the point that has still
not occurred to a lot of art and design teachers!). The notions of
"purity" and "bias" associated with the intermixture model of primary
colours (see above) reinforce this unconscious conflation. For example,
these notions would lead us to see a greenish blue or cyan as a blue that
"contains" some yellow, and so must be an imperfect primary; the
perfect primary blue would surely be a visually pure blue that is
neither greenish nor purplish.

The three historical primaries are thus a confusion of, but also in a
sense the ancestors of, both the three subtractive colourant primaries
and the four psychological primaries (Fig. 6.2.11).
Figure 6.2.11. The historical primaries as ancestors of both the modern
opponent and subtractive primaries.

"Split-primary" palettes
Figure 6.2.12. Illustration of a "double primaries palette" from The Art of Color
Mixing (M. Grumbacher Inc, 1966).

A difficulty that immediately arises from trying to use a palette

consisting of the three historical primaries is that if the red paint is
a psychologically pure red (as it is depicted and/or described in many
traditional colour theory texts including Itten's), this red is so remote
from the subtractive primary magenta that it is found to be impossible
to mix purples above a very low chroma, if at all. This difficulty, which
long ago led printers and photographers to refine their primaries and
give us modern three-colour printing, colour photographs and colour
transparencies, is instead evaded even today by many teachers of
simplistic "traditional" colour theory, by means of a so-called "split-
primary" palette consisting of a "warm" and a "cool" version of
each historical primary.
The rationale offered in a series of books by Michael Wilcox is that our
theoretical paint primaries are in fact yellow, red and blue, but all of
our actual paint primaries contain impurities of a secondary colour,
and colour mixing depends on having a quantity of this secondary
colour impurity or "bias" in common. Thus for each theoretical primary
it is necessary to have two paints, one biased in each direction, to mix
the complete hue circle. While this rationale appears to explain why
just three yellow, red and blue paints do not mix the full range of hues,
it is at a loss to explain how just three yellow, magenta and cyan paints
do.

Wilcox's explanation of how a split-primary palette works is not

consistent with a modern understanding of subtractive mixing, and is
entirely discredited. According to Wilcox, a theoretical "perfect"
primary would reflect only wavelengths of its own hue (i.e. about a
sixth of the spectrum, rather than about two-thirds of the spectrum),
and actual paint primaries reflect mainly wavelengths of their own hue,
plus a subordinate impurity of a single adjacent hue (rather than up to
two-thirds of the spectrum in large amounts). For example, Wilcox
pictures a lemon yellow paint as reflecting just the yellow part of the
spectrum with an impurity of green, and a cadmium yellow as similarly
reflecting just the yellow part with an impurity of orange. In reality, all
yellow paints reflect most of the red, orange, yellow and green parts of
the spectrum. Blue and yellow make green not because some blue and
yellow paints contain "impurities" of green, but because subtractive
mixing exposes the substantial green-wavelength reflectance that is a
major and essential contributor to the colour of all blue and yellow
paints.

As Bruce MacEvoy has shown, Wilcox's rationale can be traced back to

early nineteenth century writers, particularly Chevreul. The split-
primary strategy is explicitly recommended by Parkhurst (1898),
although the earliest specific palette that I am aware of is that
suggested by Loomis (1947, p. 163), who listed alizarin crimson,
cadmium red or vermilion, cadmium lemon, cadmium yellow, cobalt or
cerulean blue (+ viridian) and ultramarine. Of course, six such paints
in practice actually do yield an improved gamut of mixtures over three
(Fig. 6.2.13A), but this is mainly because the modern subtractive
primaries (the so-called "cool primaries") are now included. The so-
called "warm primaries" are three high-chroma pigments that extend
the gamut of the subtractive primary mixtures, but would not give a
large gamut of colour mixtures by themselves. With six pigments a
similar or greater gamut could be obtained from many other
combinations, such as a well-spaced pair of pigments near each of the
yellow, magenta and cyan primaries, for example lemon yellow and
cadmium orange, quinacridone magenta and dioxazine violet, and
phthalocyanine blue and cobalt green. Another option would be to
choose three subtractive primaries and their paint-mixing
complements, for example, yellow and ultramarine violet, cadmium
red light and phthalocyanine blue, and a quinacridone magenta and
phthalocyanine green YS. It should be emphasized however that the
importance of the true subtractive primary hues in painting is not
merely as the basis for a limited palette, but mainly because awareness
of them enables the painter to anticipate mixing paths.

The apparent value of the split-primary system is that it evades the

difficulty that a psychologically pure yellow, red and blue palette can
not mix a full range of hues. If students were taught from the start that
the optimal colourant-mixing primaries are yellow, magenta and cyan,
and confirmed for themselves that just three such paints
actually do mix a full range of hues at moderate chroma, they would
have no reason to engage with the split-primary mumbo-jumbo in the
first place. One seriously has to question of the ethics of teachers
loading up innumerable beginning art and design students, not only
with obsolete theory, but also with the expense of buying six coloured
paints when three (a) might be all they need, and (b) would relate
directly to their digital studies, instead of to a groundless, irrelevant
fabrication.

Figure 6.2.13. Mixing paths of six ApA Ferrario pigmented inks on the Munsell
hue-chroma plane, Permanent Yellow ("warm" yellow), Lemon Yellow ("cool"
yellow), Cyan Blue ("cool" blue), Ultramarine Blue ("warm" blue), Carmine Red
("cool" red), and Vermilion ("warm" red), calculated in the
program drop2color by Zsolt Kovacs. (The actual gamuts would of course be
three dimensional, and would be extended to higher chromas in some areas by
the addition of white).

One misunderstanding that is sometimes associated with the "split-

primary" palette is the idea that you can mix an entirely different set of
colours using just the three "warm" primaries or just the three "cool"
primaries. The three so-called cool primaries - a bluish red, a greenish
blue and a lemon yellow - being the three closest to the ideal
subtractive primaries, yield a relatively evenly balanced gamut.
Compared to this, the three "warm" primaries yield higher chroma
mixtures in the yellow to red range, and generally lower chroma
mixtures elsewhere. The two gamuts of course have a large core of
relatively neutral colours in common that can be mixed equally well
with either set (Fig. 6.2.13B).

1To these we could add three additional conceptions of "primaries" from Bruce
MacEvoy's site handprint.com: (1) the L, M and S "cone fundamentals", which
MacEvoy maintains are the "real" primary colours (though he is well aware that
they are not actually colours at all); (2) the quasi-opponent colours on the
orthogonal axes of the CIECAM system (yellow, vs violet blue and crimson
red vs bluish green), which MacEvoy deems to be the "modern" primaries; and
(3) the "palette primaries" by which MacEvoy means any paint included on the
artist's palette. MacEvoy's so-called "artists' primaries" (the "simple" chromatic
colours of Leonardo, Forsius and, in MacEvoy's opinion, Alberti) are identical to
Hering's opponent hues.

Modified 2nd November, 2012. Original text here and here.

<< 1 2 3 >>

6.3 PAINT-MIXING PRINCIPLES

 Mixing with white paint

 Mixing with grey paint
  Mixing with black paint
 Mixing paints of differing hues
 Mixing of transparent paints

The concept of a three-dimensional space of hue, value and chroma

provides a vital framework for practical colour mixing, but to make use
of this framework a painter must be able to anticipate the approximate
paths in colour space that paint mixtures are likely to follow.
Fortunately, despite the potential complexity of colour mixing
involving physical colourants, paint-mixing paths in practice display
patterns of considerable regularity, which can be usefully summed up
in a few simple principles. Many of these principles can be traced back
individually as far as the eighteenth century.

Mixing with white paint

(1) Mixing a dark transparent colourant paint with increasing

amounts of white paint increases the chroma of the mixture up to a
certain point, beyond which the chroma diminishes (Fig. 6.3.1A).

(2) Thinning a layer of a dark transparent colourant paint over white

paint also increases the chroma up to a certain point, beyond which
the chroma diminishes. The maximum chroma attained by this
transparent layer is higher than that attained by physical mixture
with white paint (Fig. 6.3.1B).

Figure 6.3.1. Effect of lightening permanent alizarin crimson by adding white

paint (A) and by glazing over white (B). The second method attains a higher
maximum chroma. Both processes also involve a counterclockwise shift in hue
towards magenta (right). Photographed colours plotted in YCbCr colour space
using the program ColorSpace by Philippe Colantoni. (www.couleur.org).

(3) Mixing an opaque colourant paint with increasing amounts of

white paint tends to steadily reduce the chroma as the value
increases (Fig. 6.3.2).

Figure 6.3.2. Effect of adding white to opaque paints. Unlike transparent

pigments, addition of white to an opaque pigment like cadmium red deep does
not initially increase the chroma, but steadily reduces it. This mixture also
shows a slight drift towards crimson on the CbCr plane (C). Photographed
colours plotted in YCbCr colour space using the program ColorSpace by
Philippe Colantoni. (www.couleur.org).

Mixing either transparent or opaque paints with white paint is

commonly also accompanied by a slight to dramatic shift in hue (Fig.
6.3.3A). The direction of this hue shift depends on the undertone of the
coloured pigment, which is the hue of the pigment when it is thinned
out, and so absorbs proportionately less light. For example a magenta
pigment that absorbs 10% of the red light and 50% of the blue light as a
thin film will absorb 19% of the red light (10% + 10% of 90%) and 75%
of the blue light (50% + 50% of 50%) when twice as thick, and so will
be seen as relatively bluish when thin and reddish when thick. The
most conspicuous shifts are: red > magenta > purple, and red > orange
> yellow > green; additionally some greens and blues shift a little
towards blue-green. Thus the hue shifts, if present, are generally away
from red and towards cyan, or (apart from red towards orange) in the
direction of what would generally be deemed the "cooler" hue.
(Undertones should not be confused with colour "bias" of traditional
colour theory, which is the departure of the hue of a paint from the
assumed "perfect" primary).
Figure 6.3.3. Mixing paths of various Liquitex acrylic colourant paints with
white, grey and black paints, modelled using the program drop2color by Zsolt
Kovacs. For identity of colourant paints see Fig. 6.3.6.

Mixing with grey paint

(4) Adding grey paint to a paint of low to moderate chroma and the
same greyscale value diminishes the chroma with little effect on the
value; adding grey paint to a high-chroma paint of the same value
diminishes the chroma and reduces the value more noticeably.
Figure 6.3.4. Mixing paths of Liquitex acrylic Cadmium Red Light (A) and
a Cadmium Red Light-Titanium White-Ivory Black mixture (B) with a mixed
grey (Titanium White + IvoryBlack) of the same value (C), modelled using the
program drop2color by Zsolt Kovacs

Mixing with grey paint is a commonly used strategy for reducing

chroma without changing value; when applied to high chroma paints
some adjustment is needed to counter the small darkening that occurs.
Mixing with grey is commonly also accompanied by a shift in hue (Fig.
6.3.3B), generally in the same direction as that seen in mixing with
white paint. These shifts are accentuated if the grey is somewhat
bluish, as results from mixing most black paints with white. A perfectly
neutral grey (made for example using raw umber plus a black as the
darkener) will eliminate this latter component of the shifts, but not the
hue shifts that are due to the undertones of the coloured paints. Using
a grey to reduce chroma is an important tool that keeps the painter in
control of the value of the mixture, whereas the traditional recipe of
mixing with the complementary tends to surrender control of the value
to the paint.

Mixing with black paint

(5) Adding black paint to a coloured paint mixture generally reduces

chroma more rapidly than value, drawing the mixture along a curved
path in colour space (Fig. 6.3.5). Adding extra colourant paint is
needed to restore the mixture to the straight path of a shading series.
Figure 6.3.5. Effect of progressively adding black paint to a mixture of
W&N Permanent Alizarin Crimson and Flake White (A) compared to an ideal
shading series (line of uniform saturation) from the same mixture (B).
Photographed colours plotted in YCbCr colour space using the
program ColorSpace by Philippe Colantoni. (www.couleur.org).

Uniformly coloured surfaces passing into lower levels of illumination

are represented by image colours that decrease in value and chroma
along a line of uniform saturation (Fig. 6.3.5B). Adding black
paint alone to a paint mixture typically reduces saturation as well as
chroma and value, and thus yields shadow colours that are too low in
chroma to be correctly related to the lights. The mixing paths typically
curve in towards the neutral axis, which shows that this loss of
saturation does not simply result from the fact that the black paint is
not a "true" black but a very dark grey. (I suspect that the main reason
is that the addition of light-absorbing particles reduces the average
number of times that light is reflected between pigment grains, making
the same ratio of colourant to white particles less effective in colouring
the darker mixtures).

This tendency is the reason for black's reputation for "muddying"

colours, and for the characteristic rule of simplistic colour instruction
to not use black. With a little experience however this tendency can be
easily corrected by adding some more of the colourant(s) used until
the required chroma is restored. The "colourant" in the example of Fig.
6.3.5 should be understood to mean Permanent Alizarin Crimson
alone, and not the original full-light colour (crimson plus white). If the
colourant is the full-light colour (i.e. if the latter lacks any added white
or black paint), the tendency to lose saturation is usually much less
marked than for mixtures containing white, but if necessary the
chroma can be boosted using a transparent paint of similar hue.

Strongly coloured objects remain at high saturation even at their lowest

values, which may need to be painted using a very high proportion of
colourant; pure black paint, being a very low value neutral (grey), is
unsuited to represent the lowest values of such objects.

Darkening with black may also cause a hue shift that needs correction
(Fig 6.3.3C), but this problem occurs with any darkener. The main
shifts are red > purple, and orange > yellow> green, both of which can
be corrected by dragging the hue back with a trace of a strong orange or
scarlet colourant.

Mixing paints of differing hues

Figure 6.3.6. Mixing paths of various Liquitex acrylic paints with other paints in
the range, modelled using the program drop2color by Zsolt Kovacs.

Most mixing paths of paints of differing hues do not follow straight

lines on the hue-chroma plane, and the assumption that they should -
the so-called "color wheel fallacy" - has never been an actual
component of traditional colour theory. In fact, the notion of primary
colours requires that paint-mixing paths do not follow straight lines,
but instead follow very different paths depending on the hues involved:
paints of the primary hues should mix along lines that stay high in
chroma, following lines that bow outwards on the "colour wheel", while
paints far from the primary hues should mix along lines that keep low
in chroma while passing through the primary hues, following lines that
bow inwards on the "colour wheel". These assumptions are in fact
valid, as long as we understand by "primary colour" a paint close to the
ideal subtractive (CMY) primaries, rather than to the historical
primaries red, yellow and blue. They are corollaries of the first two of
the principles set out below:

(6) Paints that are close to the ideal subtractive primaries, yellow,
magenta and cyan, make high-chroma mixtures with distant
hues (Fig. 6.3.6 A-D).

(7) Paints that are far from the yellow, magenta and cyan primaries
(e.g. cadmium red and cobalt blue) make high-chroma mixtures with
only a few adjacent hues (Fig. 6.3.6 E-G).

Paints intermediate in hue between these two groups have mixing

patterns that are intermediate in character (Fig. 6.3.6H-J).

(8) Paints close in hue to the ideal subtractive primaries have a

relatively restricted range of paint-mixing complementaries, whereas
paints far from the hue of the ideal subtractive primaries have
complements and near-complementaries of a surprisingly wide range
of hues (Fig. 6.1.6, 6.1.7, 6.3.1).

(9) The most direct paint-mixing complementary of most paints is

roughly opposite in hue on the Munsell hue circle, except close to
the Y-PB axis, where mixing paths systematically curve through
middle chroma greens.

The exception noted under principle 9 results from the fact that yellow
and blue pigments overlap in reflectance in the green part of the
spectrum (Fig.6.3.7).
Figure 6.3.7. Spectral transmission curves for two hypothetical colourants. Two
artist's pigments can be complementary in terms of the light they reflect, but far
from complementary when mixed together. After diagrams by Pope and
Luckiesh.

Figure 6.3.8 demonstrates the same effect with actual pigments: lemon
yellow and ultramarine blue give off light that is complementary (as
can also be demonstrated with spinning discs), but when physically
mixed together they make a distinctly green mixture.This drift in the
direction of green means that we need to look for mixing complements
of ultramarine blue among pigments that are more orange than lemon
yellow, and several yellow-orange pigments are in fact found to be
paint-mixing complements. Similarly we should expect the paint-
mixing complement of lemon yellow among pigments more purplish
than ultramarine blue (ultramarine violet is about right).

Figure 6.3.8. Ultramarine blue and lemon yellow, physically mixed. These
colours, though additive complementaries (white line), physically mix to make a
series of dull greenish colours. Photographed colours shown on CbCr plane and
in side view of YCbCr space. Photographed colours plotted in YCbCr colour
space using the program ColorSpace by Philippe Colantoni. (www.couleur.org).

(10) Whether they are close to or distant from the subtractive primary
hues, paints generally yield higher-chroma mixtures with paints that
are close in hue than with paints that are more distant in hue in the
same direction.

This is another traditional principle that is supported reasonably well

in practice, with occasional exceptions. Paint-mixing paths for a given
paint mostly fan out in a fairly orderly fashion, and many of the
apparent crossings of paths are connected with the mixing of pairs of
transparent paints, whose paths in the absence of white tend to pass
very close to the neutral axis, largely because the mixtures are very
dark.

(11) Mixing with a complementary-coloured paint usually causes

value to decrease as well as chroma, but at a rate that depends on the
pigments involved, and is likely to follow a shading series only up to a
point, if at all.

In the example of the paint-mixing complementaries cadmium red and

cobalt green, the paint-mixing path is close to the general direction of a
shading series for both paints for a short distance (two or three
Munsell value units) before drifting off towards the neutral axis (Figure
6.1.2A). You may also notice here that the common assumption that
when complementary paints mix, the most neutral mixture must be the
darkest, is also incorrect.

Mixing of transparent paints

(12) Opaque colours may lighten or darken a mixture, but

transparent colours always darken.

Figure 6.3.8. Effect of adding cadmium red (B) and permanent alizarin crimson
(C) to a dark grey mixed from flake white and charcoal grey. Photographed
colours plotted in YCbCr colour space using the program ColorSpace by
Philippe Colantoni. (www.couleur.org).

Transparent paints allow light to pass through without scattering; they

always darken mixtures, because they work by simply removing light.
Opaque paints scatter light; they can lighten a darker paint when
added to it, but their effect is conditional, depending as it does on the
balance of their effect of absorbing some wavelengths with their effect
as adding reflectors of the remaining wavelengths (Fig. 6.3.8).

Painting in transparent layers takes different ways of thinking about

colour manipulation to those that work for opaque paint. A paint used
transparently does not have a single Munsell notation, and glazed over
white can have any value from its body colour (very dark for most of
them) up to near white (Principle 8, above). Transparent layers are
inclined to show colour variations that follow short segments of this
path, though these variations can be eliminated if necessary by
building up to the required colour in a few light layers. The basic
technique needed is to mentally analyze each colour into a set of
subtractive "components", i.e. to assess how much of each transparent
paint is needed to arrive, whether by mixing or glazing, at the required
hue, value and chroma.

Modified 2nd November, 2012. Original text here.

<< 1 2 3 >>

Next: Part 7: Hue

PART 7. THE DIMENSION OF HUE

7.1 FROM ARISTOTLE TO NEWTON

 Hue before the hue circle

 Newton's hue system

Hue before the hue circle

The key steps of isolating hue as an independent dimension of colour,

and of representing it as a continuous circle, were first taken in Isaac
Newton's well-known colour circle in his Optics of 1704. Some earlier
colour diagrams in the shape of a circle had been drawn in the
seventeenth century (Figs 7.1.1A, B, D), and circular diagrams
displaying colours of urine had been used for medical diagnoses since
the Middle Ages (Kuehni and Schwarz, 2008, Figs 2.3, 2.4), but all of
these included black and white among the colours of the circle. So
while colour circles had existed previously, the circular dimension
of hue begins with Newton.

Figure 7.1.1. Seventeenth century diagrams of colour order systems 1 (click to

enlarge). A, B. Two-path and five-path multilinear systems illustrated by
Sigfridus Forsius (Physica [unpublished manuscript], 1611).
C. Compound (linear-multilinear-trichromatic) system by Francois D'Aguilon
from the Opticorum libri sex (1613). D. Diagram by Robert Fludd (Medicina
Catholica, 1629), formed by joining the white and black ends of a linear system
as a circle. E, F. Later compound systems based on D'Aguilon's, published by
Athanasius Kircher (Ars magna lucis et umbrae, 1646) and Johannes Zahn
(Oculus artificialis, 1685) respectively. G. Multilinear system illustrated by
Francois Glisson in his Tractatus de ventriculo et intestinis (1677). For details
of the original publications see Kuehni and Schwarz (2008).

Among earlier systems, the most influential was the linear

scale devised by Aristotle, based on the idea that colours might be
formed by the mixing in various proportions of white and black. In
his Peri Aistheseos kai Aistheton (On sense and what is sensed, c. 330
B.C.) Aristotle arranged five chromatic colours on a line between white
and black, with the lighter colours beginning with yellow close to white,
and the darker colours beginning with blue close to black (Fig. 7.1.2A).
The fundamental difficulty with such a system, obvious to us in the
light of Newton's discovery of the hue circle, is that we can proceed
from yellow to blue by two different hue series, passing through red or
through green, which makes it impossible to define a single satisfactory
hue sequence. Thus while many later authors adopted Aristotle's idea
of a linear scale between white and black, most felt at liberty to vary the
order of the intermediate hues (Fig. 7.1.2B-E).
Figure 7.1.2. Linear colour order systems. A. Aristotle, c. 330 BC, Peri
Aistheseos kai aistheton. B. Chalcidius, c. 325 AD, from his commentary on
Plato's Timaeus,. C. Bartholomaeus Anglicus, c.1245, from his encyclopaedia De
rerum naturalis (Of natural things). D. Leonardo da Vinci,c. 1500, Trattato
della pittura. E. Robert Fludd, 1629, Medicina Catholica. All diagrams based on
purely verbal descriptions, except E (after Fig. 7.1.1D). For details of the original
publications see Kuehni and Schwartz (2008, pp. 31-33, 37-38).

Beginning with the Persian philosopher Avicenna in the eleventh

century, various writers resolved this dilemna by devising what I refer
to here as multilinear systems, in which from two to seven sequences
("paths" or "families") of colours stretch between black and white.
There are no contemporary illustrations of Avicenna's simple system
(Fig. 7.1.3A) or of the more elaborate one by his 13th century
commentator Al Tusi (Fig. 7.1.3B), but the Finnish mathematician
Sigfridus Forsius (1611) used circular diagrams to illustrate a
comparable pair of systems, a two-path "ancient" system (Fig. 7.1.1A,
7.1.3D) and a five-path "right" system (Fig. 7.1.1B, 7.1.3E) respectively.
These diagrams and that of the English physician Glisson (Fig. 7.1.1G,
7.1.3F) independently show the multilinear sequences as straight and
curved lines stretching between black and white. Several writers have
interpreted the array of arcs in Forsius' second diagram as a sphere,
but neither the diagram nor the accompanying text provide evidence
that Forsius intended this (Kuehni and Schwartz, 2008, pp. 45-46). In
the simpler multilinear systems each sequence includes more than one
hue in the modern sense, but in others the sequences are
effectively tint/shade scales of a single major hue category. Tint/shade
scales are described in painter's practical manuals from the twelfth
century (Kuehni and Schwarz, 2008, p. 35), and continue to feature
prominently in later colour order systems, such as the colour "stars" of
Lacouture and Itten (Fig. 7.2.4L, 7.2.6H).

Figure 7.1.3. Multilinear colour order systems. A. Avicenna, 1015, Liber de

anima, with "paths" through pallidus (uncertain, either yellow-green or
grey), rubeus (red) and indicus (indigo blue). B. Al Tusi, 13th c. commenatary
on Avicenna. C. Alberti, 1435, Della Pictura, with "genera" passing through red,
blue, green and grey/beige (bigia et cinericia). D. Sigfridus Forsius, 1611
manuscript Physica, "ancient" system (see Fig. 7.1.1A), with colours in two
series, through red and blue. E. Forsius' second system (see Fig. 7.1.1B), with
five series through red, yellow, green, blue and grey. F. Glisson, 1677, Tractatus
de ventriculo et intestinis, (cf. Fig. 7.1.1G). Diagrams A, B and C, which were not
illustrated by their authors, are modeled on the later diagrams by Forsius and
Glisson. For details of the original publications see Kuehni and Schwarz (2008).

The elaborate diagram by the Jesuit mathematician Francois D'Aguilon

(Figs 7.1.1C, 7.1.4) ingeniously reconciles three disparate systems: (1)
a linear scale of the Aristotelian kind, (2) multilinear genera like those
of Avicenna and Forsius, and (3) the newly
ascendant trichromatic system of three primary colours - yellow, red
and blue. The latter are arranged in order of tonal value between white
and black in a linear scale of five "simple" colours. The graceful arcs in
the upper part of the diagram resemble similar arcs in medieval
musical diagrams (Kuehni, undated), but their function here is to
represent Avicenna/Forsius-style colour sequences passing between
white and black through yellow, red, blue and grey respectively. The
ancient dilemna of where to put green relative to red on the linear scale
is resolved by removing it entirely, as the compound of yellow and blue.
The two remaining compounds complete the possible pairwise
combinations of yellow, red and blue. By following the arcs of the lower
part of the diagram from yellow through gold, red, purple, blue, green,
and back to yellow, one can trace the succession of hues of the familiar
colour wheel, but this colour wheel "foetus" was to remain attached to
its linear Aristotelian parent until Newton published his hue circle in
1704. Later seventeenth century compound systems illustrated by
Kircher (Fig. 7.1.1E), Traber (Kuehni and Schwartz, 2008, Fig. 2.21)
and Zahn (Fig. 7.1.1F) are in essence identical, except that some lines
do double-duty for the superimposed trichromatic and multilinear
systems.
Figure 7.1.4. Interpretation of D'Aguilon's diagram from the Opticorum Libri
Sex (see Fig. 7.1.1C) as a compound linear-multilinear-trichromatic system.

Newton's hue system

Newton presented his circular diagram in the Opticks of 1704 as an

approximate guide to the additive mixing of coloured lights, explaining
that the colour resulting from such mixing would lie at the "center of
gravity" of the component lights, weighted according to their relative
"number of rays" (Fig. 7.1.5). He further stated that colour mixtures on
the circumference would be "intense and florid in the highest degree"
and that other colours would be intense in proportion to their distance
from the centre, thus making his diagram the first graphical
representation of the dimension of saturation, as well as hue. Newton's
hue scale, which is still widely given as the colours of the spectrum and
the rainbow, consisted of seven hues, red, orange, yellow, green,
blue (or "blew"), indigo and violet, spaced around the circle according
to the intervals of a musical scale. This scale is not one of the modern
tonal (i.e. major or minor) scales, which had become standard by the
beginning of the eighteenth century, but is a Dorian scale, one of eight
older traditional modal scales.

Figure 7.1.5. A. Colour circle from Newton's Opticks (1704), showing a

desaturated orange light mixture (z) at the "centre of gravity" of its seven
unequal spectral components, indicated by circles of differing size.
Source:Posner Library. B. Coloured version of Newton's diagram, showing the
interpretation of Newton's hue names adopted here. The spacing of the red
through violet hues is taken unchanged from my linear RGB circle (Fig. 4.3.4),
so that opposed spectral hues in the diagram are true additive complements.
Two additional discoveries are implicit in Newton's diagram: (1) the
concept, later named metamerism, that the same colour can result
from any number of different combinations of spectral components,
and (2) the relationship, later termed complementarity, that lights
having opposite colours on the circle, such as Newton created
experimentally by splitting the spectrum in various ways, will mix to
make white light (Fig. 7.1.6). Newton himself stopped short of asserting
that the latter rule applied to opposite pairs
of monochromatic lights, ostensibly because he had not succeeded
in demonstrating this result experimentally, but perhaps also because
it would have meant conceding a point to his early critic, Huygens. It is
difficult to overstate the importance of Newton's discoveries, which
contributed directly or indirectly to a huge part of our conceptual
framework of colour in science, technology and art - not to mention the
discovery of radio waves, microwaves, and the rest of the
electromagnetic spectrum! Once hue and saturation had been added to
brightness as independent dimensions, it was only a small step (first
taken by Brook Taylor in 1719) to recommend an explicitly three-
dimensional conception of colour. The hue circle and the concept of
complementarity were particularly important developments for artists,
and even the concept of "warm" and "cool" hues does not seem to have
taken hold until after artists saw their hues laid out in a circle.

Figure 7.1.6. Generalized illustration of Newton's various experiments showing

that splitting the spectrum produces coloured lights with opposite "centres of
gravity" that can be recombined to make white light.
Though Newton's circle first appears in the Opticks of 1704, the
musical comparison to which we apparently owe it is contained in De
Colorum Origine (The Origin of Colours), the second part of the Latin
manuscript of his Cambridge optical lectures deposited at that
university in 1672. Throughout this document, translated by Shapiro
(1984), Newton refers to five "principal" or "most conspicuous"
colours (like Aristotle before him, and Munsell after him): red, yellow,
green, blue and violet/purple. However when discussing a very cleanly
separated spectrum (Lecture 3) and again when discussing the mixing
of adjacent spectral colours (Lecture 8), he uses a finer scale of ten
"gradations" in which typical representatives of the five principal
colours alternate with intermediate colours (Fig. 7.1.7). He then claims
in Lecture 11 that because the most typical representatives of four of
the five principal colours are off-centre in the spaces assigned to them,
the spectrum can be subdivided more "elegantly" by giving two of the
intermediate gradations, orange and indigo, their own spaces (Fig.
7.1.7, right). Then in the following part of Lecture 11 Newton reports
the similarity between the spacing of the colours in his seven-hue
spectrum and the spacing of the stops in a Dorian scale.

Many have suggested that Newton settled on seven spectral hues

because of a preconceived attachment to his comparison with a musical
scale (which is plausible), or because of a supposed "obsession" with
the number seven (which is groundless). However, one could equally
well speculate that the common alternative of seeing six spectral hues
originates from a preconceived attachment to a symmetrical colour
wheel based on the three historical primaries. Before that idea gained
hold, Aristotle saw three or four colours in the rainbow in
the Meteorologica (Fig. 6.2.1A), and recognized five principal
chromatic colours in De Sensu, and Theodoric of Freiberg (c. 1310) saw
four rainbow colours (Kuehni and Schwarz, 2008). In any event,
Newton correctly recognized that the number of colours seen in the
spectrum was a function of how closely you looked, and depending on
the latter might be five, seven, ten or "indefinite".

Partly because the traditional "artists' colour wheel" has a place for
orange but only one blue, the spotlight of skepticism has mostly fallen
on Newton's indigo, although some relief has arrived in recent decades
in the form of the RGB(CMY) hue circle, which recognizes a deep
"blue" and cyan as separate hues. The modern meaning for "cyan" of
greenish blue dates from Helmholtz (1866), who suggested renaming
Newton's "blue" and "indigo" as "cyan-blue" and "indigo-blue"
respectively. Newton had used the Latin name cyaneus (derived from
the ancient Greek for "blue") in his ten-hue spectrum for the
presumably typical gradation of blue between the intermediate
gradations thalassinus (sea green = the green-blue gradation)
and indicus (indigo = the blue-violet gradation), but in his seven-hue
spectrum, indicus includes the blue-violet boundary (the "most perfect
indigo") plus an adjacent part of the blue of the five-hue spectrum,
while caeruleus (blue = Helmholtz's cyan-blue) extends from the most
typical blue up to the blue-green boundary (Fig. 7.1.7). Newton's indigo
might therefore be expected to include the deep, violet-tending "blue"
of modern technology (the "B" in RGB), while his blue would include
the hue range of process cyan inks up to digital cyan. This correlation is
compatible with Newton's placement in his circle of indigo as the
complement of yellow to yellow-orange, and blue as the complement of
middle red and orange, but not yellow, and also agrees with the
opinions of earlier writers including Helmholtz, von Bezold, Rood, and
Evans, who related Newton's indigo to the hue of ultramarine
(though cf. McLaren, 1985, who argued that Newton's indigo was an
indistinguishable division of violet).

Figure 7.1.7. Newton's colour terminology in the manuscript of his Cambridge

optical lectures, De Colorum Origine (1772). A, one of the passages of the latin
text giving the names of ten "gradations". B, Newton's spectral subdivisions
aligned with the reproductions of Newton's optical lecture diagrams by Shapiro
(1984).

Newton admitted in his optical lectures that a match between the

precise boundaries of the spectral intervals and those of a musical scale
could not be conclusively demonstrated from the evidence, but said
that he favoured his subdivision for two reasons: (1) "because it may
perhaps suggest analogies between harmonies of sounds and
harmonies of colours" such as are known to painters, and (2) because
"the affinity between the extreme red and violet, the ends of the
spectrum, is of the same kind as between the first and last notes of the
octave". The suggestion of "analogies between harmonies of sounds
and harmonies of colours" was not unreasonable at the time,
particularly in the context of Newton's hypothesis that colour stimuli
were conveyed to the brain by vibrations in the optic nerve.

More compelling apparent support was in store for him when he came
to arrange his colours in a circle (Fig. 7.1.8A, B). Like all modal scales,
the Dorian mode can be sounded by stopping a plucked or bowed
string at points at simple fractions of its length: in this case 8/9, 5/6,
3/4, 2/3, 3/5, 9/16 and 1/2 (Fig. 7.1.8A, blue fractions). An
approximation of these notes can be heard by playing the white notes
on a modern keyboard ascending from the note D. In introducing his
circle in the Opticks, Newton gives this scale in the form of
the intervals 1/9, 1/16, 1/10, 1/9, 1/10, 1/16, and 1/9, which refer to
the proportion of the remaining part of the string occupied by the
space to each note (Fig. 7.1.8A, red fractions). The relative sizes of
these proportional intervals therefore differ from the
absolute spacings on the string, and it is the former that Newton used
to derive the angular divisions of his circle, in line with its music theory
progenitors. The use of interval spacing rather than absolute spacing
happens to substantially improve the alignment of the additive
complementaries, for example placing middle green instead of blue-
green - the actual middle of Newton's spectrum - opposite extraspectral
"purple" (magenta), and blue-green opposite spectral red. It may have
been this serendipitous circumstance that ultimately convinced
Newton to publish his surprisingly daring analogy.
Figure 7.1.8. A. Newton's diagram of the spectrum (from Birch, 1757) aligned
with the divisions of the Dorian scale, showing the position of the stops
expressed as a fraction of the total length of the string (blue fractions), and in
terms of musical intervals, calculated as a fraction of the remaining length of
the string from the preceding stop (red fractions; the second 1/10 is mistakenly
given in most editions of the Opticks as 1/16). B. Circle divided in proportion to
the physical sizes of the divisions (blue fractions). Note that if Newton had used
this method, blue-green would have fallen opposite magenta. C. Circle divided
in proportion to the sizes of the musical intervals (red fractions), the method
used by Newton.

1Historical diagrams in Figs 7.1.1, and 7.2.1-3 sourced from Kuehni and Schwarz
(2008), Spillman (2009), various public domain sources, and the collection of
the author.

Modified April 15, 2013. Original text here.

<< 1 2 3 4 5 6 7 >>

7.2 THE RYB HUE CIRCLE OR "ARTISTS' COLOUR WHEEL"

The RYB hue circle or "artists' colour wheel" is a hue system structured
around the three historical primary colours, red, yellow and blue, and
the historical complementary relationships red-green, yellow-
violet/purple, and blue-orange. This hue system, though founded on a
scientific understanding of colour that was comprehensively
overturned in the second half of the 19th century, is still used as the
basis for colour education in many art and design classes today, even at
tertiary level.

 Origins of the "artists' colour wheel"

 What colour are yellow, red and blue?
 Itten's colour theory and hue system

Origins of the "artists' colour wheel"

Figure 7.2.1. Eighteenth century hue sytems (click to enlarge). A, B, Hand-

painted colour circles from the 1708 edition of the Traite de la peinture in
mignature. C, Louis-Bertrand Castel, 1740, L'Optique des couleurs, as
illustrated by Kemp, 1990. D, Tobias Mayer, 1758, De affinitate colorum
commentatio, as illustrated by Georg Christoph Lichtenberg in 1775. E, F,
Moses Harris, c. 1770-72, The natural system of colours. G, Ignaz
Schiffermuller, 1772, Versuch eines Farbensystems. H. Moses Harris,
1776, Exposition of English insects [1782 edn]. For details of these publications
see Kuehni and Schwarz (2008).
After Newton introduced the circular dimension of hue in
his Opticks of 1704, it was only a small step to arrange the three hues of
the seventeenth century artists' linear scale around a circle. This step
was first taken in an anonymous chapter on pastel painting added to
the 1708 Hague edition of the anonymous Traite de la Peinture en
Mignature, in a pair of hand-coloured circles showing seven and twelve
equal-sized divisions respectively (Fig. 7.2.1A, B). In the text the
anonymous author hesitates over whether there are really three
"primitive" colours or four (yellow, blue and two reds - "fire red" or
vermilion, and crimson ), but interestingly the primary status of "red"
ultimately survives the fact that it had to be mixed here from two
pigments. The seven-hue circle shows the four pure pigments, plus
three mixtures respectively of yellow, blue and the reds, while the
twelve-hue version adds further mixtures, to place yellow, the mixed
"primary" red and blue evenly spaced around the circle. In both
diagrams the clockwise sequence of hues follows the order of the
seventeenth century linear scale (yellow-red-blue), and so is reversed
compared to Newton's spectral order (red-yellow-blue).

This twelve-hue circle is the earliest example of the so-called "artists'

colour wheel", an arrangement of regularly spaced hue divisions
structured around the three historical primaries. While it incorporated
Newton's discovery that hues form a closed loop, the "artists' colour
wheel" was otherwise an implicit rejection of the assumption that
Newton's circle and rules of additive mixing (in which all spectral hues
are "primary") also applied to paints. Despite this early example,
published circular systems are few in number until the first decades of
the nineteenth century, and in this early period triangular systems such
as those of Mayer (1757) and Sowerby (1809) are just as prominent.
(These triangular systems also arrange hues in a closed loop, but
designate colours by the proportions of their yellow, red, and blue
"components", rather than by hue as such). Nineteenth century "colour
wheels" are very diverse geometrically, and incorporate variously
subdivided triangles, hexagons, and 6- to 24-pointed stars, either in
combination with a circle, or alone (Figs 7.2.2, 7.2.3, 7.2.4).

Several of the systems shown here follow Moses Harris (c. 1770-72) in
consisting of two or even three diagrams, in order to display hue
categories specifically for low-chroma colours, designated
as tertiary colours by Field (1817). These low-chroma yellows, reds and
blues were respectively known as olive, brown and slate (following
Harris) or citrine, russet and olive (following Field). This use of the
word "tertiary" persists today alongside an entirely different usage,
which can be traced back at least to Ruskin (1877), for third-order hues
located between adjacent primary and secondary hues. In the context
of the assumptions of traditional colour theory, tertiary colours of the
first kind "contain" all three historical primaries, while those of the
second kind "contain" two primaries in unequal proportions.

When our anonymous author of 1708 wrapped the primary hues of the
linear scale symmetrically around a circle, the "secondary" or
"composite" hues, now designated purple, green and orange, were each
placed equidistant between their two "component" primaries, and
therefore directly opposite the third primary. As in our naming of the
historical primaries, the unconscious influence of the psychological
primaries can be seen in our choice, out of the continuous sequence of
hues obtained by mixing yellow and blue paint, of the simple name
"green" for the colour automatically placed opposite "red". In the
1708 text, no significance whatsoever was attached to pairs of opposing
colours, beyond a general recommendation that colours distant on the
circle should not be mixed. However when Harris published his colour
circles in c. 1770-72, opposing colours on his circle were claimed to
reveal colours of maximum visual contrast, colours of
afterimages, and colourant-mixing complements. This assumption of
the "all-purpose" nature of complementary pairs remains typical of
simplistic traditional colour theory today.

During the late nineteenth century revolution in our understanding of

colour, it progressively became evident that several fundamentally
different kinds of opposing relationships exist among hues, including
opponent relationships (Section 7.3), additive complements (Section
7.4), and colourant-mixing complements (Section 7.5), and that the
hues that oppose each other are somewhat different in each case. In
reality we need somewhat different hue circles to represent each set of
relationships accurately. Which one we use depends on what kind of
question is being asked. After this revolution, authors who continued to
use the historical primaries mostly felt obliged to justify their position,
given that the red, green and blue/violet primaries of Helmholtz and
Maxwell were now widely regarded as the "true" primary colours by
scientists. Their standard argument ever since has been that artists
must work with pigments, and therefore their colour classification
must be constructed in terms of the mixing of pigments rather than
light. This argument neglects three crucial points:

1. Artists may also be said to work with the light reflected by their
artworks, and when the question concerns this visual stimulus (as in
simultaneous contrast), the appropriate framework is one based on
light mixture.
2. Most fundamentally, artists work with the perceptions of their
audience, and hue perceptions are structured around the four
psychological primaries.

3. The clarification of the theoretical basis of subtractive mixing had

revealed that the optimal primaries for colourant mixing were not in
fact yellow, red and blue, but yellow, magenta and cyan, the
complements of the additive primaries.

What colour are yellow, red and blue?

The "artists' colour wheel" can now be seen as an unconscious

compromise, developed on the assumption that only a single hue circle
was needed, at a time when the different kinds of opposite or
complementary relationships were not understood. While they are
unanimous in designating the primary colours as "yellow", "red" and
"blue", such diagrams show far less unanimity in the paint or ink
colours used to illustrate these primaries (e.g. Figs 7.2.2, 7.2.3, 7.2.4).
This variation reflects the nature of the historical primaries as a
conflation of the four psychological primaries - yellow (Y), red (R), blue
(B) and green (G) - and the three subtractive primaries, yellow (Y),
magenta (BR) and cyan (GB), additionally complicated by the relative
chroma of the available pigments. Thus, nineteenth century "primary"
reds range from the the bluish red pigments closest to the ideal
subtractive magenta primary (insect, madder or alizarin crimsons, all
around Munsell 2.5R) through psychologically middle reds (around
5R) to the highest-chroma available red (vermilion, around 7.5R).
Nineteenth century "primary" blues range from the best available
subtractive primary (Prussian blue, around 7.5B-10B) and
psychologically middle blue (around 10B) to the highest-chroma
available blue (ultramarine, around 5PB). Nineteenth century
"primary" yellows span the range of the then available high-chroma
yellow colourants (roughly 7.5YR-5Y), which includes middle yellow
(around 5Y). In the twentieth century the invention of phthalocyanine
and quinacridone pigments respectively expanded the ranges of
"primary" blues and reds closer to the ideal subtractive primaries.
Figure 7.2.2. Selected nineteenth century hue systems, part 1 (click to enlarge).
A, B, Baumgartner and Muller, 1803, Esthetique de la Toilette. C, Sowerby,
1809, A new elucidation of colours. D, Runge, 1809, in Runge,
1840, Hinterlassene Schriften. F-G, Runge, 1810, Farben-Kugel. H, Goethe,
1810, Zur farbenlehre. I, Gregoire's c.1810-1820 system, from Gregoire,
1820, Tables des couleurs sur trois feuilles. J, Hayter, 1813, An introduction to
perspective, drawing, and painting. K, Klotz, 1816, Gruendliche Farbenlehre.
Figure 7.2.3. Selected nineteenth century hue systems, part 2 (click to enlarge).
A, Merimee, in Brisseau-Mirbel, 1815, Elemens de physiologie vegetale et de
botanique. B, Field, 1817, Chromatics [1st edn]. C, D, Hayter, 1826, A new
practical treatise on the three primitive colours. E, Merimee, 1830, De la
peinture a l'huile. F, Hay, 1836, The laws of harmonious colouring [3rd edn]. G,
de Lisle, 1838, Chromagraphie. H. Merimee, 1839, The art of painting in oil,
and in fresco. I, Field, 1841, Chromatography [2nd edn]. J, Delaistre,
1842, Cours methodique du dessin et de la peinture. K, Forbes, 1849, On the
classification of colours. L, Field, 1850, Rudiments of the painter's art.
Figure 7.2.4. Selected nineteenth century hue systems, part 3 (click to enlarge).
A, Adams, 1860, Die Farben-Harmonie in ihrer Anwendung auf die
Damentoilette. B, Chevreul, 1861, Expose d'un moyen de definir et de nommer
les couleurs. C-F, Schreiber, 1868, Die farbenlehre. G, Bacon, 1872, The theory
of colouring. H, Blanc, 1876, Grammaire des Arts du Dessin. I, J, Davidson, in
Field & Davidson, 1874, A grammar of colouring, applied to decorative
painting. K, Ridgway, 1886, A nomenclature of colors for naturalists. L,
Lacouture, 1890, Repertoire chromatique. M, Sanford, copyright 1891, in
Sanford, 1910, Sanford's manual of color. N, Earhart, 1892, The color printer.
O, Prang system, in Prang, Clark and Hicks, 1893, Suggestions for a course of
instruction in color for public schools. P, Milton Bradley system, in Maycock,
1896, A class book of color.

The need for a system of fixed standards for hue terms such as
"yellow", "red" and "blue" was recognized as early as 1810 by Gaspard
Gregoire, who published an atlas of 1,351 coloured samples arranged
according to hue (Fig. 7.2.2I), value and relative chroma, but the work
apparently had little impact, and no copies survive today (Kuehni and
Schwarz, 2008, pp. 82-3). Two colour classifications published in the
United States in the late 19th century by the printer Louis Prang and
the games manufacturer Milton Bradley were also structured around
the traditional primary and secondary hues, although in Bradley's
system these were known as the six spectral primaries. Prang's system
involved scales of twelve hues (designated R, RO, O, YO, Y, etc.) and 24
hues (R, RRO, RO, ORO, O, etc), while Bradley's system had 18 hues
(R, OR, RO, O, etc.) for high chroma or "pure spectrum" colours and 12
hues (R, OR, O, YO, Y, etc.) for low chroma or "broken spectrum"
colours (Fig. 7.2.4). Denman Ross (1907) used Prang's 12-hue system
of hue designation for his own colour order system, and Albert Munsell
used a similar lettering code for the five-hue system of his Atlas of
1915.

Figure 7.2.5. A, B, Standard colour chips representing the Prang system, from
Prang, Clark and Hicks, 1893, Suggestions for a course of instruction in color
for public schools. C, Colour chips of the Bradley system, from Bradley,
1895, Elementary color [3rd Edn]. The chips in A to C are glued-in painted
cards, as in Munsell's Atlas. Collection of the author.

Itten's colour theory and hue system

The "artists' colour wheel" has persisted alongside more sophisticated

hue systems, and is still widely taught in art and design courses today.
This persistence, and the corresponding exclusion of modern colour
theory, are associated more than anything else with the extraordinary
influence of one book, The Art of Color (1961) by Johannes Itten
(1888-1967). Both the abridged 1970 edition by Faber Birren and the
mass-produced second edition of 1973 have been widely used as
textbooks, and many later texts and websites recycle the approach to
colour theory presented in Itten's book. Itten's reputation derives
ultimately from his association with the German Bauhaus, which has
been enormously influential in art and design education since the mid-
20th century. Itten was responsible for implementing the widely copied
Preliminary Course (Vorlehre) at the Bauhaus, and taught there for
three and a half years until the "mystical" elements in his teaching, the
disunity caused by his continuing efforts to "rule the Bauhaus", and his
habit of coming into the craft workshops giving instructions on
technical matters he did not understand (Forgacs, 1995) lost him the
support of some of the staff and students, including the director,
Walter Gropius.

Colour was taught at the Bauhaus by artists successively including

Itten, Ludwig Hirschfeld-Mack, Paul Klee, and Wassily Kandinsky, as
well as by various masters in the specialized workshops such as wall-
painting, weaving, etc. (Gage, 1993, pp. 259-263; Poling, 1973). The
dominant influence on colour theory initially was the former
mentor/teacher of Itten and Hirschfeld-Mack, the painter Adolf
Hoelzel (1853-1934), who also devised virtually the entire structure of
the Preliminary Course implemented by Itten (Parris, 1979). Hoelzel's
approach to colour was relatively broadly based, and although he
attached prime importance to Goethe, he also incorporated a wide
range of other influences including the scientists Wilhelm von Bezold
(1837-1907), Ogden Rood (1831-1902) and Wilhelm Ostwald (1853-
1932). Hoelzel used different colour diagrams for different purposes,
six of which were printed in a 1919 publication by the Pelikan Company
(Fig. 7.2.6A-F). The first four of these are structured around the
historical primaries, secondaries and tertiaries, and are derived from a
colour theory manual of 1868 by Guido Schreiber (Fig. 7.2.4C-F). They
consist of two symmetrical six-hue colour circles showing the primaries
and secondaries, and secondaries and tertiaries respectively (Fig.
7.2.6A, B, cf. Fig. 7.2.4C-D), a circle divided into six unequal hue
sectors calculated from the ratios of the "proportional activity of the
retina" that Schopenhauer (1788-1860) had asserted to be the basis of
colour perception (Fig. 7.2.6C, cf. Fig. 7.2.4E), and a triangle, often
known today as the "Goethe triangle" and incorrectly attributed to
Goethe himself, but actually devised by Schreiber on the basis of the
systems of Tobias Mayer and George Field (Fig. 7.2.6D, cf. Fig. 7.2.4F).
The fifth and sixth diagrams are circles using more modern hue
systems devised by Bezold (Fig. 7.2.6E, F). These "diatonic" (8-hue)
and "chromatic" (12-hue) circles formed the main framework for
Hoelzel's schemes of colour harmony, which included complementary,
split-complementary, triadic, tetradic and "irrational" schemes. In the
12-hue circle (Fig. 7.2.6F, G), based on hue divisions Bezold had
regarded as perceptually equidistant, an equilateral triad connects the
modern subtractive primaries ("purple"1, cyan-blue and yellow),
leaving the historical primaries unequally spaced. Klee and Kandinsky
emphasized the historical primaries/secondaries in their teaching and,
like Hoelzel, held Goethe's colour writings in extraordinarily high
regard, but Kandinsky also incorporated the work of scientists
including Helmholtz and Ostwald. Outside their classes the main
emphasis was placed on the practical aspects of colour in relation to
the materials being used, and the colour system that had greatest
currency was that of Ostwald (Poling, 1973).
Figure 7.2.6. A-F, Six colour theory diagrams by Hoelzel from Wollner,
1919, Farbkreisen zu den Vortragen von Adolf Holzel, Pelikan Co., Stuttgart. G,
12-hue or "Chromatic" colour wheel from Hoelzel, 1919, Einiges uber die Farbe
in ihrer bildharmonischen Bedeutung und Ausnutzung, with English
translation of German colour names. H. Itten's color star of 1921 (lithograph on
paper), with English translations of the German colour names. The hue
sequence is essentially that of Hoelzel's 12-hue circle, but is reoriented to place
Itten's preferred division between "warm" and "cool" hues, at the yellow/yellow-
green and blue-violet/purple-violet boundaries, vertically. The very uneven hue
steps in the red-blue sector might indicate alteration of some of the colours.

In The Art of Color of 1961 Itten still derived much of his teaching from
Hoelzel's theories of colour harmony and contrast, and even
appropriated some of Hoelzel's diagrams, including the circle based on
Schopenhauer's ratios, which Itten mistakenly thought were Goethe's
(Itten, 1961, pp. 104-5). However, while explicitly critical only of
Ostwald, Itten's post-Bauhaus book also eliminates almost all of the
other elements Hoelzel had derived from late 19th and early 20th
century science. Thus although his "colour star" lithograph of 1921 (Fig.
7.2.6H), which he had used as the basis of his teaching of colour at the
Bauhaus (Itten, 1975, p. 33), follows Hoelzel's 12-hue system derived
from Bezold, in the colour star and other diagrams in The Art of
Color Itten reverts to a system comprising evenly-spaced historical
primaries and secondaries, with six intermediates. (Additionally, he
flipped the hue sequence vertically, reversing the hue order and placing
primary yellow at the top centre). This post-Bauhaus version of Itten's
star is thus only a variation on its 19th century forerunners such as that
of Lacouture (Fig. 7.2.4L). For a three-dimensional framework Itten
(like Klee) rejected the Ostwald system used fairly widely in the
Bauhaus and simply ignored the Munsell system, reverting instead to a
simple spherical model externally like that of Runge (1810), but
internally subdivided in a manner suggested by Brucke (1866). Itten's
sphere places the strongest colours of all hues on the equator, ignoring
their different value and absolute chroma, and thus lacks a consistent
representation of the dimension of value that is vital to most painters.

The printed colours used to illustrate the 12 hues of Itten's post-

Bauhaus system vary between different editions, and even between
different diagrams in the same copy, but his three primary colours are
defined in the text as perceptually pure hues, oddly enough using the
names of the four psychological primaries of modern science, which
are otherwise completely ignored in the book: "a red that is neither
bluish nor yellowish; a yellow that is neither greenish nor reddish;
and a blue that is neither greenish nor reddish" (Itten, 1961, p. 34;
emphasis mine). Itten's secondary colours, designated orange, green
and violet, are also defined perceptually, as not leaning towards either
primary "component", but are claimed to be also exactly
complementary to the third primary in paint-mixing, in simultaneous
and successive contrast, and even in expressive interpretation (Itten,
1961, p. 137). Itten designates as tertiary six third-order hues located
between adjacent primary and secondary hues; these are given
compound names consisting of the primary followed by the secondary,
e.g. yellow-orange, red-orange, etc., as in Prang's 12-hue system.
Mainly because it is structured symmetrically around just three of the
four psychological primaries, Itten's hue circle is uneven perceptually,
with relatively large hue steps in the yellow-green-blue sector.

Many of Itten's explanations of colour theory are as retrograde as his

classification. His explanations of successive and simultaneous
contrast echo Goethe's invocation of a mysterious "eye animism" ("the
eye demands the complement..."), like Goethe completely ignoring the
hypothesis of photoreceptor adaptation already suggested by Palmer
and Young. His discussion of colour mixing in paints is framed in
terms typifying the intermixture model that had been rendered finally
obsolete by Helmholtz. No mention is made of the subtractive
primaries cyan and magenta, or indeed of subtractive mixing, or of the
fact that Itten's primaries are so remote from ideal cyan and
(especially) magenta that paints of these hues make a poor set of
colourant-mixing primaries, quite unsuitable for mixing purples. Itten
is also said to have "vastly simplified" Hoelzel's system of seven
contrasts of colour in his own version of that system (Parris, 1979, p.
99).

The Art of Color embodies an approach to colour theory and

classification fixated at a stage at least seventy years before the
founding of the Bauhaus, and fabulously anachronistic at the present
time. In several ways it is the colour theory equivalent of so-called
"creation science". Whatever one's opinion of other aspects of the
Hoelzel-Itten innovations in art and design education, the teaching in
many institutions of a near monoculture of Itten's simplistic post-
Bauhaus colour theory has been disastrous for the general level of
public understanding of colour today.
1Hoelzel's
"purple" sits between "carmine" and "purplish-violet", and so must be
near modern process magenta in hue.

Modified April 15, 2013. Original text here and here.

<< 1 2 3 4 5 6 7 >>

7.3 HUE SYSTEMS BASED ON OPPONENT-HUE

RELATIONSHIPS

 Introduction
 Goethe's hue system
 Hering and Ostwald
 The NCS system
 An opponent-hue "artists' colour wheel"
Figure 7.3.1. An opponent-hue "artists' colour wheel", showing the hue
categories currently in wide use among painters arranged in a full opponent-hue
framework with the green psychological primary included (see below). This
arrangement restores the underyling symmetry of hue perceptions that is lost in
the traditional "artists' colour wheel". C,M,Y = best available colourant-mixing
primaries.

Introduction

According to the opponent theory proposed by the German

physiologist Ewald Hering (1834-1918), all hues we experience are
combinations of adjacent pairs of the four psychological primary or
unique hues, red, yellow, green and blue, and no hues are
combinations of yellow and blue, or red and green. In the widely
accepted modern zone theory, which reconciles the opponent theory
with the three-receptor theory of Young and Helmholtz, these unique
hues originate as red vs green and yellow vs blue signals created by the
brain indirectly from the relative responses of the three cone cell
types. Thus while the number three is the key to questions involving
colour stimulus (additive and subtractive mixing), four is the key to
colour experience, and there is no simple relationship between these
two kinds of "primary colour". The three historical primaries, red,
yellow and blue, seem to represent an unconscious confusion of the
two concepts (section 6).
An opponent-hue system is structured around the four psychological
primaries and their opponent relationships. We automatically use an
opponent-hue system as a frame of reference in everyday speech: when
we say that a colour is a slightly yellowish green or greenish yellow, we
are describing that colour in relation to an implicitly recognized unique
hue. Notice that we never say "bluish yellow", even if we have been
thoroughly indoctrinated with the historical primaries! Opposing hues
in opponent-hue systems are opposite experiences having no
perceptual components in common. Perhaps counterintuitively, the
four unique hues are not equally spaced perceptually (see Fig. 7.3.8B
below), and while unique yellow and unique blue are also near
opposites in both additive and perceptually-spaced systems, unique red
and unique green are not.

In addition to pure opponent-hue systems, a number of quasi-

opponent systems have been devised in which the primaries or chroma
axes coincide roughly but not exactly with the four psychological
primaries. The historically important Ostwald system was strongly
influenced by Hering, but also incorporated additive complementary
relationships in its hue circle, which is structured around four
primaries known as yellow, red, blue, and "seagreen". Later quasi-
opponent hue systems are seen is the additive CIE L*u*v* system and
the perceptually-spaced CIE L*a*b* system and CIECAM02 colour
appearance model. In the Lab colour space used in Photoshop,
the a axis represents crimson (+) vs green (-) chroma and the b axis
represents yellow (+) vs blue (-) chroma.

An opponent-hue framework is especially relevant for questions

involving our mental experience of colour, such as arise in the
interrelated fields of:

1. colour psychology: studies testing the mental and physiological

effects of colours.
2. colour symbolism: investigations of the symbolic use of colour in
different historical and cultural contexts.
3. colour expression: theories of the "meaning" of colours, often
purporting to draw to some extent on studies of the first two
types.

Goethe's hue system

Given that the four psychological primaries are the basic building
blocks of our experience of hue, it is not surprising that they turn up
repeatedly in colour order systems long before their explicit
recognition by Hering and his contemporaries. In the middle ages
Theodoric of Freiberg saw these four hues as the colours of the rainbow
(Kuehni and Schwarz, 2008, p. 36), and the same four appear
alongside white and black as the simple colours in the systems of
Leonardo (Fig. 7.1.2D) and Forsius (Fig. 7.1.3E) (section 7.1). (They
have also been seen by some in the system of Alberti, though it seems
more likely that his four simple colours were intended to be red, green,
blue and grey). Credit for a major step towards the concept of
opponency must go to Goethe, whose theory of the origin of colours is
based on an explicitly opponent relationship of the yellow and blue
primaries (Fig. 7.3.2A). For a very concise summary of the essence of
this theory please see this paragraph.

Hooke and Huygens had previously suggested that yellow and blue
were the two fundamental primary colours, but Goethe also came close
to recognizing red and green as pure or simple colours, despite
believing the latter to be a mixture of yellow and blue, and his theory
indirectly contrasts pure red (his purpur), seen as the result of
"augmentation" or reddening of both blue and yellow, with middle
green, seen in his theory as a balanced, simple mixture of the two (Fig.
7.3.2B).

Figure 7.3.2. Quotes from Goethe's writings concerning yellow, blue, red and
green.

One might say that despite his theory of the origin of colour, which
itself has no physical or psychological validity, our innate opponent
framework of colour perception made its presence felt in Goethe's
choice of hue terms to describe that theory. This opponent framework
emerged most openly in his section on colour terminology, where he
proposed a fully opponent system of 12 hue names formed as
combinations of the four symmetrically arranged hues red, yellow,
green and blue, in which no hue names are combinations of yellow and
blue, or red and green (Fig. 7.3.3A, B). Goethe himself pointed out that
this four-hue system is suggested by the German language itself.

Figure 7.3.3. Goethe's 12-step opponent system of hue terminology (B) and its
relationship to his three afterimage "complemental " pairs (C), indicated with
red arrows in both.

Goethe's unique foreshadowing of opponency is sometimes confused

with his discussion of afterimage complementary ("complemental")
colours, which were already known in the eighteenth century. Goethe
himself must have recognized that two different kinds of opposite
relationship were involved, as two of his complemental pairs diverge
from his primordial opposition of yellow vs blue (Fig. 7.3.3B,C). [In
Goethe's theory these oppositions would be considered subjective and
objective respectively]. The third complemental pair, recorded as
"purpur"-green, on the other hand was made to align with the red-
green axis by using the term "purpur" with two distinct meanings.
Unique red and unique green are not additive or afterimage
complements - the complement of middle green is magenta and vice
versa. "Purpur" literally means "purple", and is the name Goethe used
for the magenta afterimage-complement of "balanced" green, and also
for a magenta colour appearing in edge spectra that, in modern terms,
is the additive complement of spectral green (Fig. 7.3.4). However
"purpur" also has an important place in Goethe's theory as the ultimate
degree of the "augmentation" or reddening of both blue-red and
yellow-red, and in this role is necessarily as well as explicitly described
as an absolutely pure red, neither yellowish nor bluish (Figure 7.3.2B).
Consistently with the latter, it is also described as being the colour of
fine carmine, intermediate between the slightly bluish and slightly
yellowish varieties of the colourant (Goethe, Theory of Colours, 796,
799-800). Goethe used "purpur" as a "weasel word" that could
be defined as pure red (about Munsell 5R), but could be stretched
where needed to include the distinctly bluish (magenta) afterimage and
additive complements of green (roughly Munsell 10P to 5RP).
Figure 7.3.4. A. Goethe's illustrations of the edge spectra seen when a white
band on a black background (left) and a black band on a white background
(right) are observed through a glass prism from various distances (Theory of
Colours, 243-6, pl. 2). In Goethe's view the green in the middle of the lower two
light spectra is formed by mixing of the yellow and blue fringes seen separately
in the upper view (see explanation to Fig. 6.2.6). The magenta colour labelled
"purpur" in the middle of the corresponding 'dark spectra' is in modern terms
the additive complement of this green. B. Simulation of Goethe's experiment.

Hering and Ostwald

Figure 7.3.5. A, Hue circle and B, diagrammatic hue page showing a colour
specified in terms of black (b), white (w) and full colour (r) components
(Hering, Outline of a Theory of Light Sense, tr. Hurvich and Jameson, 1964). C.
Hering's six fundamental colours arranged in a distorted octahedron, from
Ebbinghaus' Psychology, an elementary text-book (1908), translated from his
German text of 1897.

In the original opponent theory of Ewald Hering, white and black,

together with red, yellow, green and blue, were regarded as the six
fundamental colours of visual perception. In Hering's view, all colours
could be considered as mixtures of full colour (pure hue), white and
black components (Fig. 7.3.5B). This was a major departure from most
other systems, which instead used relative or absolute lightness or
brightness as an important dimension. Hering did not attempt to
provide colour samples arranged according to his system, apart from
his purely illustrative hue circle (Fig. 7.3.5A, Fig. 6.2.9).

As Hering himself noted, several other 19th century psychologists had

already settled on red, yellow, green and blue as the four fundamental
hues, and the four-hue system began to appear frequently in
psychology texts from around the turn of the century, including those
of both Hofler and Ebbinghaus in 1897 (Fig. 7.3.5C) and Titchener in
1901 (Kuehni and Schwarz, 2008, pp. 101-3). Nevertheless, Hering's
theory was still often presented as a less likely alternative to the
trichromatic theory of Helmholtz and others until the idea that the two
could be combined in some form of zone theory eventually gained
general acceptance, particularly through the efforts of Leo Hurvich and
Dorothea Jameson in the 1950's. The four psychological primaries are
now explicitly acknowledged in the official definition of the word "hue"
by the Commission Internationale de L'Eclairage (CIE) as "the
attribute of a visual sensation according to which an area appears to be
similar to one of the perceived colours, red, yellow, green and blue, or a
combination of two of them".
Figure 7.3.6. Ostwald system. A, Ostwald's four principal and four intermediate
hues from his Die Farbenfibel (1917) . B, Analysis of a colour into white (d),
black (e) and full colour (f) components, from his Einfuhrung in die
Farbenlehre Vol 2 (1919). C, model of the Ostwald system showing double-cone
form (wikipedia). D, Blue/yellow and seagreen/red hue pages from Ostwald's
colour atlas, Der Farbkorper (1919).

Wilhelm Ostwald (1853-1932) published the first atlas of colour

samples influenced by Hering's opponent theory. Ostwald followed
Hering in classifying colour in terms of white/black/full-colour
content, and structured his hue circle around four principal hues, but
in order to make opposing colours additive complements, the place of
the green psychological primary was taken by a bluish green known as
"sea green". The Ostwald system was widely used in art education in
the middle decades of the twentieth century, was the official system for
all levels of art education in Great Britain between the wars (Fig.
7.3.7A), and was promoted worldwide in many popular books on colour
for artists by authors such as Faber Birren. The system was gradually
abandoned in the scientific community beginning in the 1940's, and by
the time the quasi-opponent CIE L*a*b* and CIE L*u*v* systems were
developed in the 1970's, the tide had turned strongly against the
inclusion of techincal elements in general in art education. In recent
decades the three primaries of "traditional" colour theory have become
so prevalent in the education systems of most countries that many
artists today assume that they have been used universally for centuries.
Figure 7.3.7. A. Colour plate from Colour Practice in Schools by O.J. Tonks,
Winsor and Newton, London, 1934. B. An Ostwald watercolour box, by Winsor
and Newton, c. 1930's. C. The Color Harmony Manual, a spectacular colour
atlas based on the Ostwald system, sold by the Container Corporation of
America from 1946 to 1972.

The NCS system

Figure 7.3.8. A. Opponent-hue circle from Johansson's Den allmanna
farglarans grunder (1937). B, Hue circle from Hesselgren's
Fargatlas (1953), structured around the psychological primaries, but
also expressing perceptual hue spacing by placing varying numbers of
hues in the four quadrants.

A different Hering-inspired approach was suggested in 1937 by the

Swedish physicist Tryggve Johansson (1905-1960). Johansson
designed an opponent hue circle (Fig. 7.3.8A) that could be used in
three alternative systems in combination with dimensions of (1)
chroma and lightness, (2) saturation and lightness, and (3) chroma and
"cleanness" (a converse of black content). The Swedish architect Sven
Hesselgren (1907-1993) produced an atlas of colour samples using the
second of these proposals, but although Hesselgren's hue circle was
likewise structured around purely opponent hue axes, it also expressed
perceptually-equal spacing by placing varying numbers of hue divisions
in the four quadrants (Fig. 7.3.8B). Later both Johansson and
Heselgren contributed to the development of the Swedish NCS system,
in which both the hue circle and the hue pages are structured in line
with Hering's system (Kuehni and Schwarz, 2008).

Figure 7.3.9. A, Elementary colours of the NCS System. B, Symmetrical

double cone form of the NCS System. C, Hue circle of the NCS
System. D, Y10R colour triangle, showing location of the colour 2050
Y90R. Source: http://www.ncscolour.com.

The Swedish Natural Colour System (NCS), first published as an atlas

in 1978, is an important modern colour order system of the pure
opponent-hue type, in which colours of each hue are classified
according to Hering's system of pure hue/white/black content. The
system is embodied in a commercially available atlas and other
products in which Hering's opponent hues are displayed as specific
"elementary" colours (Figure 7.3.9A). Hue is specified in terms of
percentage steps around the NCS hue circle clockwise from the
preceding elementary hue (Figure 7.3.9C). Variations ("nuances") in
each hue are depicted on a triangular diagram in terms of white, black
and coloured components, and specified in decimal amounts of black
(S or "blackness") and of colour (C or "chromaticness") (Figure 7.3.9D).
The third, "whiteness" component makes up the remainder and so does
not need to be specified. The resulting system has the form of a
symmetrical double cone (Figure 7.3.9B) in which all full colours are
located on the equator. The system does not represent the dimension of
lightness (value), which limits its usefulness for most painters, but it
nevertheless enjoys wide popularity as a reference system for
architects, designers and paint manufacturers (Kuehni and Schwarz,
2008, p. 110), and is the national standard in Spain, Sweden and
Norway.

An opponent-hue "artists' colour wheel"

The four psychological primaries and their opponent relationships are

central to our modern understanding of colour vision, but are ignored
in most accounts of "traditional" colour theory for artists. Such
accounts, however, commonly define or use the three historical
primaries in the same way that the psychological primaries are defined,
as perceptually pure hues, (e.g. in Itten's colour theory), and when
painters think of their paint hues in terms of "warm" and "cool" reds,
yellows and blues, they are actually using an incomplete opponent-hue
framework, with the green psychological primary missing. This
omission distorts the psychological opposition of yellow vs blue, and
destroys the underlying symmetry of hue perceptions. Figure 7.3.1 (top
of page) shows widely used painters' hue terms arranged in a full
opponent-hue framework with the green psychological primary
included. This arrangement offers some improvement over the hue
spacing of the traditional colour wheel, restores the true psychological
oppositions and the symmetry of hue perceptions, and clearly
distinguishes the four psychological primaries from the three
colourant-mixing primaries. In doing so it of course does not display
exactly equal perceptual spacing, or colourant-mixing
complementaries.

There is a good consensus among painters that a middle red lies

between cadmium red (warm) and alizarin or quinacridone crimson
(cool), middle yellow lies between cadmium yellow medium (warm)
and lemon yellow (cool), and middle green lies between
phthalocyanine green YS (warm) and phthalocyanine green BS (cool).
These painters' conceptions of the psychological primaries lie at about
5R, 5Y and 5G respectively on the Munsell hue scale. There is also a
good consensus that middle blue lies at about the hue of cobalt blue
(Munsell 5PB), between the hues of ultramarine blue and
phthlocyanine blue, although there are two opposed traditions
regarding the warm-cool polarity of this pair. What I take to be the
majority view regards ultramarine as the warm blue, and in Fig. 7.3.1
would align the warm-cool axis with the orange/blue-green direction,
and place the warm-cool transition/ boundary/ no-man's land at right
angles to this at purple/yellow-green.

Modified July 28, 2013. Original text here.

<< 1 2 3 4 5 6 >>

HUE CIRCLES BASED ON ADDITIVE COMPLEMENTARIES

A hue circle based on additive complimentary relationships is the

relevant choice for all questions where light stimulus is the issue. As
there are many kinds of questions where this is relevant, the additive
hue circle is particularly important for artists. Some examples include:

 Results of optical mixing, as in pointillist painting, operate on

additive (partitive) principles. Even though we are dealing here
with artists paints, we are seeing the result of additive mixing
of light from these paints, not subtractive mixing. So, for
example, interspersed dots of ultramarine and yellow paint (in
the right proportion) will make grey, not the green you would
expect from a pigment mixing wheel, and would get if you
physically mixed or glazed the same pigments.
 The coloured afterimage seen after exposing the eye to coloured
light for an interval of time is essentially a temporary illusion of
light composed of the wavelengths "missing" from the stimulus,
and so is the additive complementary. Consequently successive
contrast, which is the tendency of these afterimages to influence
the apparent colour of other objects, also goes to the additive
complementary.
  Hue shifts due to simultaneous contrast, including the
complementary colours seen in the shadows of coloured lights,
also go to the additive complementary.
 Questions of colour harmony, if viewed from the point of view of
mutual enhancement of colours, depend on simultaneous and
successive contrast, and so should be worked out using the
additive hue circle. Ogden Rood (1879) argued that colours
within 80 or 90 degrees of each other on such a wheel exhibited
negative interactions, and in doing so gave an explanation for the
effectiveness of colour harmonies based either on
complementary pairs or on equally spaced triads (Figure 7.8).

Figure 7.8. "Chromatic Circle displaced by Contrast, showing the effects

produced by red on the other colours" from Rood (1879). The arrangement of
colours was determined by experiments with spinning discs, and places colours
opposite their additive complements.

 In colour photography, both conventional and digital, colour

correction uses the relationships of the additive hue circle. For
example, excessive blueness is countered by adjusting the
strength of the additive complementary with a yellow filter.
 In digital painting programmes, the usual conceptual
framework for hue is the Hue angle of HSB and HLS colour
space, which is expressed graphically as the RGBCMY hue circle
or hexagon (Figure 7.9). This system is based on additive
complementary relationships, with the three screen primaries R
(scarlet or orange-red), G (yellowish green) and B (deep violet-
blue) placed arbitrarily at 120o degrees to each other, and
 opposite their additive complementaries Y (yellow), M (magenta
or red-violet) and C (cyan or blue green).

Figure 7.9. THE RGB-CMY hue circle.

In this system the Hue angle is measured clockwise from "Monitor

Red" at 0 degrees. For the fully saturated colours, each 60 degree
difference in hue angle between a primary and an adjacent secondary
marks a 255 unit change in one of the RGB components (Table 7.1),
and so each degree of hue angle represents a change of 255/60 or 4.25
units. The hue angle of desaturated colours is obtained by treating the
colour as having a white and a coloured component, and measuring the
hue of the coloured component by the same method.

Colour name RGB components Hue Angle

Red R 255 0

Yellow R 255 G 255 60

Green G 255 120

Cyan G 255 B 255 180

Blue B 255 240

Magenta R 255 B255 300

As always, there is not an exact equivalence between the psychological

concept defined by experience and a psychophysical concept defined by
the stimulus. Most observers see tints of H = 0 as more crimson and of
H = 240 as more violet than the pure colour, and shades of H = 60 as
more greenish and H = 300 as more purplish than the pure colour
(Figure 7.10).

Figure 7.10. Variations in perceived hue within colours of identical Hue Angle.
Within each horizontal row, all squares have the same "Hue angle", the same
R/G/B ratio, and therefore presumably the same dominant wavelength.

In the original conception of the Munsell system (Munsell, 1905), the

hue circle was based on additive complementary relationships, in that
pairs of opposite colours were chosen that mixed to neutral grey in
experiments with spinning discs. Munsell (1905) also regarded his
opposite colours, consistently with this, as visual complements.
Munsell's thoroughly decimal system specified five principle colours
(red [R], yellow [Y], green [G], blue [B], and purple [P]) and five
intermediate colours (YR, GY, GB, PB and RP), all equally spaced.
Munsell subsequently divided each of these ten principle and
intermediate hues into ten clockwise steps, e.g. 5 R is the hue in the
middle of the range of the principle hue red.

Adjustments based on extensive colourimetric measurements in the

1940's resulted in what is known as the renotated Munsell System,
which is widely used today as a means of specifying colour in science
and industry, as well as among a relatively small but committed band
of painters. The renotated system allows conversion between Munsell
and CIE specifications (such as L*ab), and thereby to approximate
RGB values (or precise values if the particular RGB space is specified).
The three pairs of screen primaries and secondaries red-cyan, green-
magenta, and yellow- blue, which are exact additive complements in
any specified RGB space, return Munsell hues that are roughly
opposite in the renotated system, though not exactly so in the case of
yellow and blue (Figure 7.11). In addition, the angular spacings
between the three pairs in the Munsell system differ noticeably from
those of the RGB-CMY hue circle, but not in a way that has an obvious
superiority in perceptually equal spacing. These differences between
the two hue circles are of relatively small practical importance,
however; both the RGB-CMY and the Munsell hue circles make
suitable conceptual frameworks for artists in situations when additive
complementary relationships are relevant.
Figure 7.11. Munsell hue circle, with indicative positions of the RGB primaries
and secondaries shown by coloured dots. The Munsell hues were obtained
using Wallkillcolor's Munsell Conversion Software for the colours at 80%
brightness (to avoid artefacts that arise in the conversion of some RGB colours
with values near 255), assuming illuminant C, 2 o observer. The actual values
returned were: Red (R 204) = 8.03R; Yellow (R 204 G 204) = 0.56GY; Green (G
204) = 9.71GY; Cyan (G 204 B204) = 8.05BG; Blue (B 204) = 6.68PB; Magenta
(R 204 B 204) = 8.19 P.

<< 1 2 3 4 5 6 7 >>

HUE CIRCLES BASED ON PIGMENT-MIXING

COMPLEMENTARIES

We noted previously that pigmentary mixing differs from ideal

subtractive mixing in three main ways:
 1. ideal pigment primaries do not exist
2. our best pigmentary primaries differ in hue from ideal magenta
and cyan, and
3. the complementary relationships are different, especially in the
yellow-blue direction

Quinacridone magenta/"Permanent Rose", Pthalocyanine blue (green

shade), and any bright yellow such as a "Lemon Yellow" mix about the
greatest gamut that can be obtained from just three artists' paints
(Figure 7.12).

Figure 7.12. Mixtures of Art Spectrum Lemon Yellow (A), Quinacridone

Magenta (B), and Pthalocyanine Blue GS (Green Shade) (C) on a white
ground, showing general extent of gamut.

Without doubt the main reason why most artists look at a colour wheel
is as a guide to the mixing of paints. Strictly speaking, no precise hue
circle showing pigmentary complements can be drawn up, because the
results of subtractive mixing of actual colourants depend on the details
of their absorption curves, and can not be predicted exactly from
visual inspection of hues. Nevertheless, most artists will prefer a
tolerably accurate diagram to a precise table. One option is to base a
pigment mixing circle on our actual best primary colours, which we
may as well place symmetrically, and opposite their actual pigmentary
complementaries (Figure 7.13). Phthalo blue GS can be thought of as
the pigmentary complement of "scarlet", meaning the orange extreme
of red, between cadmium scarlet and cadmium orange. Quinacridone
magenta, has for its pigmentary complement pthalocyanine green
(yellow shade). Neutralizing yellow paints generally requires a carefully
balanced mix of blue and violet or magenta paints. Ultramarine violet
is produced by heating ultramarine blue and contains a residue of the
latter in varying amounts in different brands; the bluer variants may
work as a mixing complement for yellow paints.

Figure 7.13. A simple conceptual layout of hues for pigment mixing based on
our best pigmentary primaries and their complements.

In placing the range of reds opposite the interval from green and blue,
this arrangement is more accurate than the conventional artists colour
wheel, which seems to have been influenced here by
the psychological opponency of red and green. Stephen Quiller has
already published what is essentially this arrangement in his
book Color Choices. This hue circle strictly applies only for questions
involving the subtractive interaction of artists paints, which includes
both physical mixing and glazing of paints, but not optical mixing.
Interestingly however, afterimage complementaries show a similar
pattern, in that they also tend to be offset from the additive
complementary near the yellow-blue axis.

<< 1 2 3 4 5 6 7 >>
ORTHOGONAL SYSTEMS

The dimension of hue is not our only option for describing colour. For
many scientific purposes, systems are used that dispense with hue and
chroma, and instead use orthogonal coordinates. Two types of systems
of orthogonal coordinates, CIE Lab space, and a series of related colour
spaces used for video systems, will be mentioned briefly here.

Most books on colour science and many websites give good accounts of
CIE Lab space, and the details and history of its derivation from
colourimetric data from large numbers of individuals (for example see
the account at the site by efg's Computer Lab). While
the specification of Lab colour eliminates the dimension of hue,
its graphical representation in the Adobe Colour Picker in practice
offers the digital painter a means of selecting colours according to the
dimensions of hue, relative chroma and lightness (Figure 7.14).

Figure 7.14. CIE Lab Colour space. Left: Gamut of RGB colours in Lab space
viewed in ab plane, using ColorSpace. Note that the additive complimentary
pairs are not exactly opposite each other in CIE Lab space. Right: Graphical
representation of Lab space in the colour picker in Photoshop CS2, showing
relationship to hue, relative chroma and lightness.

Bruce MacEvoy has published colour wheels showing representative

colourimetric measurements of a large number of watercolour
pigments plotted using in a modified version of the CIE Lab system,
and more subsequently in the more recent CIECAM system (link). As
colour is measured in these systems in colorimetric units, unknown
colours or mixtures strictly speaking can not be plotted on these
diagrams by artists lacking spectrophotometers, although positions
could of course be guessed roughly based on the plotted pigments. The
opposing hues or "visual complementaries" in the CIE Lab system are
not exact additive, pigmentary or psychological complements.

YUV, YIQ, and YCbCr colour spaces, devised for video systems, also use
orthogonal coordinates instead of hue (Figure 7.15). YCbCr, which I
have been using for the illustrations of image colours in space
throughout this site, is a transformation of YUV that conveniently
results in the RGB gamut, when viewed from above, forming a regular
hexagon with the screen primaries evenly spaced and opposite their
additive complementaries, as in the RGB-CMY hue circle (Figure 7.15).

Figure 7.15. Plan views of RGB gamut in (A) YUV, (B) YIQ and (C) YCbCR
colour spaces. In YCbCr the screen primaries are evenly spaced, as in the RGB-
CMY hue circle, though in the reverse order.

<< 1 2 3 4 5 6 7 >>

WARM AND COOL HUES

Figure 7.17. Two of the earliest representations of cool-warm directions on a
colour wheel. Left: "Cold" and "warm" colours from Charles
Hayter's Perspective (1813). Right: "Hot" (upper right) and "cold" (lower left)
directions from George Field's Chromatography (1835).

The familiar expressions warm and cool refer to psychological

associations of colours rather than any physical properties of the light
inducing them. They seem not to have been applied to colours until
soon after artists first saw their range of hues laid out in a circle.
[Charles Hayter's colour circle of 1813 (Figure 7.17) appears to mark
their earliest appearance in a published colour system, but the terms
can be traced back in artist's correspondence as far as 1727 (Gage,
1999, p. 22)]. The introduction of the terms may simply reflect the fact
that, when laid out in this way, it is immediately evident that the range
of hues associated with sunlight (yellow to crimson) are confined to
one half of the circle, and sit opposite a range of hues centred on blue
or blue-green.

The expressions warm and cool are all too often used by artists in a
vague sense that fails to separate the concepts of hue and chroma. For
example, a teacher may tell a student that "that red area needs to be
warmer". This could mean either that the hue is correct but the chroma
is too low, or that the chroma is correct but the hue needs to shift
towards orange. Either way it probably means that the teacher is not
thinking clearly in terms of the three dimensions of colour.
The terms warm and cool can however play a useful role, as long as
they are always used in a precise sense referring specifically to relative
hue. In this clearly preferable sense, the terms provide a useful means
for referring to relative positions and directions around the hue circle.

Since "warm" and "cool" are psychological associations, it is not

surprising that there is a great deal of inconsistency in usage. At the
extremes some authors, such as Arthur Pope, regard yellow and violet
as marginal hues, neither warm nor cool, while others regard yellow as
the warmest and violet as the coolest hues respectively. However the
commonest positions in the literature seem to be to regard red, orange-
red or orange as the warmest hue, and blue or blue-green as the
coolest.

If a warm-cool polarity is to be used to describe relative positions

around the hue circle, some consistent and rationally justifiable
definition would be helpful. Since it is a psychological association that
we are talking about here, it may be relevant to note that both the
yellow-blue and the red-green opponent pairs have a warm-cool
polarity that is unambiguous judged by the criteria of most authors.
This could be taken to support a decision to define the warmest and
coolest colours as yellow-red (orange) and blue-green respectively, a
position close to the average usage. These two hues are of course only
directly opposite each other on the psychological colour wheel.

Figure 7.18 Suggested relative warm and cool directions on the three basic
colour wheels.

The tendency of warm hues to appear to advance and cool hues to

recede has been given a simple and accurate explanation in terms of
the physical phenomenon of differential refraction (Luke, 2001). This
phenomenon causes the light rays from a red object to focus further
from the lens of the eye than the light from an equidistant blue object,
at a point where light rays from a slightly closer object would focus.
Though not our primary means of depth perception, this effect can give
a strong impression that the red object is closer, particularly if it is not
contradicted by other sources of depth information (Figure 7.19).

Figure 7-19. Demonstration of advancing and receding colours. Notice that it is

not the colour with the greatest brightness (green) that seems to advance most,
as has sometimes been asserted, but the one associated with the longest
wavelength (red)

<< 1 2 3 4 5 6 7 >>

Next: Part 8: Lightness and Chroma

PART 8. LIGHTNESS AND CHROMA

THE DIMENSION OF LIGHTNESS

Lightness is technically defined as the perceived brightness of an object

compared to that of a perfect white object (Kuehni, 2005). Lightness
refers specifically to object colours, not colours seen as independent
lights, and ranges between black and white through the various shades
of grey.

The lightness of a light-reflecting surface depends on the proportion of

light energy reflected from a surface, but the relationship to the
amount of light energy (radiance) is a nonlinear one, i.e. lightness
is not directly proportional to radiance. A middle grey surface that
looks visually halfway between white and black reflects only about 18%
of the light energy reflected by a white surface. Charles Poynton has
emphasized the importance of distinguishing beween between linear
and nonlinear scales. Linear scales, such as radiance and CIE
luminance, called Y (= radiance weighted to spectral sensitivity of a
human observer), are proportional to light energy. Nonlinear scales,
such as CIE lightness (L) and video luma (confusingly, also
designated Y and often referred to as luminance), are intended to be
proportional to the scale of human perception. Digital artists will know
of CIE lightness (L) from its utilization in Lab colour space at the core
of programmes such as Photoshop.

CIE lightness is normally measured on a scale of 100, and is derived

from CIE luminance by a modified cube root relationship (Yn=
luminance of white):

L = 116 (Y/Yn)1/3 - 16; 0.008856 < Y/Yn

CIE lightness is a psychophysical scale based on colourimetric

measurements, and may not quite coincide with lightness as
experienced by an observer, even under ideal conditions. In particular,
certain colours have a tendency to look lighter than a grey of the same
CIE lightness, an effect known as the Helmholtz–Kohlrausch
Effect (Figure 8.1). In Photoshop, this can result in some small
surprises when converting images to greyscale mode, which translates
all colours to a grey of the same value of CIE lightness.
Figure 8.1. Helmholtz–Kohlrausch Effect. A: Various colours on a grey
background, all measuring L = 50 in Photoshop. B: same image converted to
greyscale mode.

Artists painting in traditional mediums generally use other scales of

lightness. Many painters use the system promoted by Denman Ross
(1907), consisting of nine tonal levels from black to white (inclusive).
Such a scale can be created by mixing a medium grey visually halfway
between white and black, a light and a dark grey halfway between
medium grey and white and black respectively, and intermediate greys
in the four intervals so created. Ross designated these levels by a
system of letters: Blk, LD, D, HD ,M, LLt, Lt, HLt, and Wt. The system
of seventeen levels that would result from adding the next generation
of intermediates is too unwieldy as a basic scale, although of course
refined tonal painting ultimately requires discrimination of half steps
and smaller intervals on the nine-level scale.
Figure 8.2. Lightness scales of Denman Ross and Albert Munsell. A, Lightness
("value") and hue ("color") scales of Denman Ross, showing location of highest
chroma (color-intensity") versions of each hue in bold (Ross, 1907). B, Value
scale of Munsell, from Cleland (1921).

When physically creating a lightness scale in this way one soon

becomes aware that the perception of relative lightness is rather
strongly dependent on background lightness (Figure 8.3). This effect of
background lightness is called "crispening", and is a specific instance of
simultaneous contrast, which tends to intensify lightness differences
either side of the greyscale value of the background. The degree of
illumination in which the scale is viewed also seems to strongly affect
the apparent spacing of the values.
Figure 8.3. Effect of background on perceived lightness. This lightness scale
created in Photoshop may not be perfectly even, but where is the biggest jump?
The phenomenon of "crispening" makes lightness intervals close to the
background lightness look relatively great.

Albert Munsell described a perceptually uniform lightness ("value")

scale as a part of his invention of the first quantitative classification of
colour in terms of three perceptual dimensions. Munsell (1905)
described a scale of eleven levels (and thus ten intervals), from black at
zero to white at ten. In his Atlas of the Munsell Color System (1915)
this was interpreted as a nine level scale of actual paint between
"unattainable" black and "unattainable white on one and ten
respectively (Cleland, 1921). The resulting scale, with actual black paint
on one and white paint on nine (Figure 8.2B), was effectively identical
to that of Ross. In the modern Munsell system, most black paints have
a value of about 2 (though a glossy black paint can be as low as 0.5),
and most white paint has a value of about 9 (Luke, 2001). Other
painters however, including the influential American teacher Frank
Reilly, have expanded Munsell's scale to eleven levels of paint values,
with black and white paint on zero and ten respectively.

A vigorous band of predominantly tonal realist painters active today

are strongly committed to the use of the Munsell System as a
framework for seeing and mixing colour. These artists in effect train
themselves to see the hue, brightness and colorfulness of the light
coming from their subjects in terms of the Munsell hue, value and
chroma of the paint that they will use in their paintings. Other painters,
however, while thinking of colour in terms of
the conceptual dimensions of hue, lightness, and chroma, do not use
the specific framework of Munsell units. Whatever one decides
regarding the need to use Munsell's specific scales of hue and chroma,
the use of some sort of absolute scale of lightness, whether that of Ross,
Munsell, or Reilly, is of unquestionable value for the purpose of judging
tones in painting.

<< 1 2 3 >>

THE DIMENSION OF CHROMA

Chroma is the perceived strength of a surface colour, the degree of

visual difference from a neutral grey of the same lightness. In recent
technical literature of colour appearance, chroma is defined
as "colorfulness" of an object relative to the brightness of a white object
similarly illuminated, which allows for the fact that a surface of a given
chroma displays increasing "colorfulness" as the level of illumination
increases (Figure 8.4).

Figure 8.4. Chroma and colorfulness. A surface of a given chroma is more

"colorful" in higher illumination (B,D) than in low illumination (A,C). Ttonal
painters would observe this difference in brightness and "colorfulness" of light,
and represent it with paint areas of different lightness and chroma, in order to
create the illusion of a surface of uniform chroma under varying light. A and D are
in fact exactly the same screen colour, but are seen in normal viewing mode as a
light colour in shadow and a dark colour in light respectively. (As always,
squinting helps to see this relationship).

The term chroma was invented by Munsell (1905), who subsequently

quantified the term as an open-ended scale of perceptually uniform
steps, defined in relation to his atlas of colour samples (Munsell, 1915).
In doing so he showed that the maximum chroma actually attainable in
surface paints was different for different hues. As new pigments are
discovered the range of possible chroma for each hue increases .
Chroma ranges beyond 20 for some normal reflecting materials, as is
as high as 30 for some fluorescent paints.

As well as being measurable in perceptually equal steps, as in the

Munsell system, chroma can also be described on a relative scale
between zero and the maximum possible for the hue, as is implicit in
the circular shape of the traditional artist's colour wheel. Denman Ross
(1907) suggested using a relative scale of one-quarter, half, three-
quarter and full "color-intensity".

We saw in the introduction that in order to have high chroma, a surface

must reflect light of relatively high saturation and brightness for a
given degree of illumination. For any given hue, it is possible to
imagine a surface that reflects all wavelengths of light that contribute
to the sensation of that hue, and none that cause desaturation of that
hue. Such a surface would have the maximum possible chroma for that
hue: any surface of higher lightness necessarily reflects less saturated
light, and any surface that of lower lightness obviously reflects less
bright light. Conceptually the range of possible variations of a single
hue may therefore be visualized as a triangle, in which the range of
possible chroma becomes progressively more restricted as one
approaches white and black respectively. The lightness level at which
this maximum chroma occurs is highest for yellow and lowest for hues
around violet-blue (Figure 8.5).
Figure 8.5. Diagram of all possible variations of a single hue of red, showing
variations in the range of possible chroma with lightness.

In the digital realm, RGB colours conform precisely to this simple

triangular geometry (Figure 8.5), but actual surface colours also
conform closely to this general triangular pattern (Figure 8.6).
Figure 8.6. Range of chroma for values 0-9 for each of the ten principal hues of
the Munsell system. White (value 10) omitted from each diagram. Pages
generated using WallKillColor Munsell Conversion programme 7.0.1.
<< 1 2 3 >>

HUE-CHROMA-LIGHTNESS COLOUR SPACES

A colour wheel represents chroma on a radial axis from the centre, and
hue by position around the wheel, but a third dimension representing
lightness is necessary if all colours are to be represented. The simplest
way to do this is to add this third axis at right angles to the colour
wheel, creating a solid such as a sphere or a symmetrical double cone.
[A cylinder would also be possible, but would not represent the visual
convergence of colours as they approach black and white respectively].
The earliest illustration of a definite colour space of this type is the
colour sphere of Otto Runge, published in 1810, and better known to
many artists in the recycled version published ed by Johannes Itten
(Figure 8.6). The idea of a double cone was used in the colour
classification by Ostwald (Figure 8.7).

Figure 8.7 . Colour sphere of Johannes Itten.

Figure 8.8. Symmetrical double cone colour solid by Ostwald.

The problem with both the sphere and the symmetrical cone
conceptions of colour space is that, as we have just seen, different hues
reach their maximum chroma at different tonal levels. Putting all of
the pure colours on the equator of the solid ensures that the vertical
dimension does not represent lightness. Consequently neither the
Runge-Itten sphere nor the Ostwald double cone is a true hue-chroma-
lightness space. If the vertical dimension of the solid is to represent
lightness, then we need in some way to tilt the colour wheel through
space, so that yellow occupies a high position opposite light grey and
blue occupies a low position opposite dark grey.

This requirement can be satisfied most simply in a skewed double

cone, a solution first suggested by Kirschman (1896) (Figure 8.9A).
Arthur Pope (1922, 1931) described in detail an essentially similar
double cone space, implicit in, though not actually illustrated in, the
colour system of Denman Ross (Figure 8.9C). Both of these double
cone solids are simple conceptual models in which chroma is
normalized, so that they have a simple circular appearance in plan
view. Following Ross, the Pope solid uses a hue circle based on the
traditional artist's colour wheel, but analogous solids based on hue
circles of psychological, additive or pigmentary complementaries could
easily be visualized. A simple conceptual model of this sort is extremely
valuable for many purposes of practical painting.
Figure 8.9. Representations of hue-chroma-lightness space by (A) Kirschman
(1896) (flipped), (B) Munsell (1915), and (C) Pope (1922), and YCbCr spaces, all
viewed from similar angles.

The Munsell system uses conceptually similar dimensions, but has a

more complex form because it is a physical system showing the
absolute variations of actual paint samples. The Munsell solid has an
irregular, tree-like appearance, reflecting the fact that the maximum
chroma manufactured paints varies markedly for different hues.
Figure 8.10. Two views of the Munsell colour space generated using a
programme by John
Kopplin (http://www.codeproject.com/directx/d3dmunsell.asp)

Screen (RGB) colours have a more complex geometry for a largely

different reason. Among full-chroma screen colours, the three
secondaries are all lighter than their adjacent primaries: L = 98 for
yellow, 60 for magenta and 91 for cyan, compared to 88 for green, 54
for red and 30 for blue. These differences can be accounted for by the
fact that in the secondary colours, pixels of both the adjacent
primaries are glowing. Magenta and cyan consequently disturb the
steady fall in lightness from yellow to violet-blue on both sides of the
hue circle (Figure 8.9D, 8.10).

Figure 8.11 . RGB colours arranged in YCbCr colour space, using the program
RGB Cube by Philippe Colantoni.
Figure 8.12 . RGB colours arranged in Lab colour space, using the program
RGB Cube by
Philippe Colantoni.

The same up and down movement might be expected in surface

colours, but is hardly evident in artists paints because our pigments are
so remote from ideal magenta and cyan (Figure 8.13).
Figure 8.13. Colours of about 100 coloured pigments from a Winsor and
Newton colour chart (pdf version).

Please note that throughout this section I have been referring to all of
these systems as colour spaces, because that is how I recommend that
painters think of them - as three-dimensional spaces through which the
artist manouvres. Arthur Pope in particular has demonstrated in detail
how a simple geometric space such as his double cone model can make
an excellent mental framework for visualizing and understanding
colour relationships. In the context of serious colour science however
the term colour space is restricted to quantitative systems that can be
mathematically transformed, and that systems that fail to meet this
criterion are referred to as colour models.
<< 1 2 3 >>

Next: Part 9: Brightness and Saturation

PART 9. THE DIMENSIONS OF

BRIGHTNESS, SATURATION AND
"COLORFULNESS"

Figure 9.1. Same saturation, different chroma and "colorfulness". All four
screen areas A-D emit light of the same saturation (pure red), but they differ
among themselves in chroma, both when seen as surfaces in the subject (A[=B]
> C[=D]) and, in a different way, when seen as surface colours in the image (B >
A[=D] > C). Light from these four areas, though of the same saturation, exhibits
progressively more "colorfulness" in proportion to its brightness.

We saw in the introduction that a different set of dimensions applies to

the visual appearance of light, as opposed to surfaces. Brightness is
the perceived intensity of a light, and saturation is the perceived
purity of colour or relative colour intensity of a
light. "Colorfulness" - the absolute colour intensity of a light
stimulus - is a function of brightness and saturation. Brightness is the
perceptual correlative of the physical parameter of luminance, which in
turn is the amount of light energy (radiance) weighted according to the
relative sensitivity to each wavelength of human vision. Saturation is
the perceptual correlative of physical parameter of spectral purity.
Colours making up an image, which can be described in terms of
lightness and chroma if looked at as surface colours, can also be
described in terms of brightness, saturation and "colorfulness" if
viewed as light coming from the image.

In earlier literature, brightness and saturation are often treated as

essentially subjective parameters, unsuited to quantification, but more
recent literature on colour appearance models is developing ways of
treating these dimensions quantitatively (Moroney et al., 2002).
However, absolute quantitative measures of these parameters, and of
their physical correlatives, are not generally used by painters, and for
most purposes they do not need to be. Digital artists are able to directly
manipulate brightness and saturation values in their images in
programmes such as Photoshop. This opens up enormous possibilities
for emulating the effects of light from the imagination, as long as
artists understand the basic principles of colour. Even so, for most such
purposes, digital artists need ony concern themselves with the specific
measures of brightness and saturation used in these programmes
(defined in relation to the gamut of available colours), rather than with
absolute quantitative measures of brightness, saturation, and their
physical correlatives.

Tonal realist painters do systematically judge the brightness and

colour intensity of the light coming to their eyes from their subjects,
but in general they do not think in terms of absolute measures of these
parameters, but only with relationships of these parameters between
the different components of their subject. They typically (and often
unconsciously) think of these relationships of brightness and
"colorfulness" in terms of the value and chroma of the paint
mixtures that they will use, sometimes in relation to an absolute
framework such as the Munsell System.

<< 1 2 3 4 >>

BRIGHTNESS-BASED COLOUR SPACES 1: RGB, CMY AND

CMYK SPACES
Figure 9.2. RGB and CMY colour space. (A) RGB and (B) CMY colour spaces,
illustrated using the programme RGBCube by Philippe Colantoni.

RGB SPACE

Since the colours on a computer monitor or television are produced

from various combinations of R, G and B components, it follows that
they can be represented by a system of three orthogonal axes
representing these components. This results in a cubic volume
enclosing all possible screen colours, with black at the origin and white
at the opposite corner. Easygoing artists like ourselves can refer to this
as RGB colour space, but when mixing in strict colourimetric circles,
take care to refer to it as the RGB colour model, which can be
embodied in various defined colour spaces, such as sRGB or
AdobeRGB.

The R,G and B components are usually reported on a scale of 0 to 255,

but can also be reported on a scale of 0 to 1 (distinguished as r,g and b
here). Confusingly, these RGB values sometimes refer to linear units
of light energy, or radiance, and sometimes to nonlinear units of
perceived brightness (i.e. in equal perceptual steps). Often no care is
taken to show which of the two kind of units is being used - you need to
work it out in each context (link). In Photoshop, both relative
brightness (B) and the R,G and B components are reported in
nonlinear (brightness) units.The conversion is:

(nonlinear) brightness = linear "brightness" 0.45

The formula has a generally similar effect to the more elaborate

nonlinear conversion between CIE luminance (Y) and Lightness (L),
 L = 116 (Y/Yn)1/3 - 16; 0.008856 < Y/Yn

CMY SPACE

In CMY space, the same RGB colours are considered as subtractive

mixtures of varying quantities of cyan (C), magenta (M) and yellow (Y)
colourants. The resulting cubic space is identical to RGB space, apart
from the fact that the origin of the C, M and Y axes is at the point
representing white instead of black (Figure 9.2B). The conversion is
given by the formulae C = 255 minus R (or 1 - r), M = 255 minus G (or 1
- g), and Y = 255 - B (or 1 minus b) respectively (Figure 9.3). Note that
C, M and Y in these formulae refer to the ideal subtractive primaries,
not actual cyan, magenta and yellow inks, i.e. C,M and Y behave as
ideal subtractive colourants complementary to the particular R,G and B
additive primaries that are in use.
Figure 9.3. Ideal conversion of RGB to CMY.

CMYK SPACE

Although cyan, magenta and yellow inks might be expected be

sufficient for colour printing, most actual colour printing uses black ink
in addition. This is partly because a mixture of the first three inks may
not yield a black that is neutral enough, or dark enough, but also
because the use of black spares the use of the more expensive coloured
inks, and also reduces the total amount of ink used, thus speeding
drying times. Conversely, it permits the use of coloured inks with better
colour rendering properties than would be possible if it was necessary
that these mix to make a dense black by themselves. The practical need
for a black component was recognized right from the invention of
colour printing by the German artist J.C. Le Blon in the early 1700's.
After Le Blon's death his former pupil, Jacques Gautier D'Agoty, in
order to protect his own patent for the four-colour process, disputed
the claims of Le Blon's workshop that the master had ever used more
than three colours. Le Blon 's supporters replied that their master kept
quiet about his use of the fourth plate because he used it in spite of
himself, and felt that it would dishonour his system (Gage, 1999, p.
139).

The amount of the black component needed is conventionally specified

by the letter K. CMYK values are typically reported as
percentages. Ideal CMYK values can be calculated by simple formulae
directly from CMY values, but these values are not accurate for colour
printing (Ford and Roberts, 1998):

Black (K) = minimum of C,M,Y

Cyan = (C - K)/(1 - K)
CMYK

Magenta = (M - K)/(1 - K)
CMYK

Yellow = (Y - K)/(1 - K)
CMYK

These "cheap and nasty" formulae for CMYK in effect divide the CMY
values into a black component (determined by the minumum value
among C, M and Y), and the relativeproportions of C, M and Y within
the remaining coloured component (Figure 9.4). In ideal CMYK, one of
the C, M or Y values is therefore always zero.
Figure 9.4. Ideal conversion of CMY to CMYK. Though this idealized
conversion is only indicative, it at least suggests how the use of black ink can
permit the same result to be obtained using less coloured ink, and less ink
overall, than with three coloured inks alone.

The conversion to CMYK values given by the colour picker in

Photoshop does not use these simple formulae, but is a much more
sophisticated colour management transformation (via Lab space) that
takes account of the colour profiles of the monitor display and printer
inks specified by the user.
<< 1 2 3 4 >>

BRIGHTNESS-BASED COLOUR SPACES 2: HSB (=HSV)

Although all screen colours can be produced by varying the R,G and B
components, graphics programmes offer alternative means of adjusting
these components that are intended to be more intuitive. HSB (=HSV),
HSL (=HLS) and HSI are three such spaces devised for this purpose.
All three are designed to resemble the system of hue, lightness and
chroma familiar to artists, but all three lack a true lightness or chroma
dimension. Of the three, HSL is perhaps the most intuitive for colour
selection, but HSB is incomparably more powerful for applying
the principles of colour, because its parameters named saturation (S)
and brightness (B) relate closely to important parameters of colours
seen as light. However both S and B have specific meanings in HSB
that differ from absolute brightness and saturation, and relate instead
to the range of possible values in RGB space. Both parameters are
given on a scale of 1 to 100.

RELATIVE BRIGHTNESS (B)

The parameter called B or "Brightness" in particular means something

quite different in HSB space to absolute brightness, and will be
referred to here as relative brightness. B measures the brightness of a
colour compared to the maximum possible for a colour of the same
hue and saturation, which means having the same ratio of R/G/B.
The numerical value of B is given by the brightness of the brightest
RGB component as a percentage of 255. Thus all colours having at least
one RGB component equal to 255 are said to have a brightness of 100.
Such colours include white, all pure colours, and
all tints (intermediates between white and a pure colour). These
colours are the brightest possible version of their particular RGB ratio;
they form the ceiling of hue-lightness-chroma colour solids such as CIE
Lab or YCbCr (link). These colours range enormously in lightness, from
L=100 for white down to a minimum of L = 30 for "Monitor Blue"
(Figure 9.5).
Figure 9.5. The top row of colours are all at maximum relative
brightness (B=100), because in each case they are the brightest possible
version of that pure colour. They vary greatly in lightness, however, as is
confirmed by their measured greyscale value (L).

We might expect that for neutral colours the relative brightness of

the light coming from an area of the screen surface, measured as a
fraction of the maximum possible brightness for the device (R255 G255
B255), should be numerically equal to the perceived lightness of
that surface, since this is also measured relative to maximum screen
brightness. Measures of both lightness (L) and relative brightness (B)
appear in the colour picker of Photoshop, and do move roughly in step
with each other for neutral colours (Table 9.1). The numbers are not
exactly equal, however, because of the different formulae used for the
nonlinear transformation of each (see RGB Space).

B 0 10 20 30 40 50 60 70 80 90 100

L 0 9 22 33 43 53 63 73 82 91 100

Table 9.1. Comparison of relative brightness (B) and lightness (L) values from
the Adobe Colour Picker for neutral colours between black (L=0) and white
(L=100).

RELATIVE SATURATION (S)

In HSB space Saturation ("S") similarly refers to saturation compared

to the maximum possible in RGB space. It is quantified by in effect
considering a colour to be divided into a coloured component and a
white component, and measures the proportion of the coloured
component to the whole (Figure 9.4). Colours are at the maximum
saturation (100) when one or two of the RGB values are zero. Fully
saturated colours occupy the outward- and downward-facing planes of
a colour space such as CIE Lab and YCbCr.

Figure 9.6. Calculation of Saturation (S) and Brightness (B) for the light green
colour R 102 G 255 B153. The total amount of light may be thought of as being
split into a white and a coloured component; S is the proportion of the coloured
component of the total. Brightness (B) for this colour is 100, because it is the
brightest possible colour with this ratio of R/G/B.

HSB space is conventionally visualized as an inverted cone or

hexagonal pyramid, in which the vertical axis represents brightness,
and the angle of divergence from the vertical axis represents
saturation (Figure 9.7). Because the vertical axis is relative brightness,
all of the pure colours, tints, and white appear at the same level, on the
the top plane of the solid. If we transform the hexacone into a space
where lightness is the vertical dimension. the pure colours take their
places at their respective tonal levels, and we get a volume of a kind
that we have already seen - in YCbCr. (A similar though slightly
different solid would result if the vertical dimension was CIE lightness
rather than Y or luma).
Figure 9.7. Relationship of HSB hexacone to the hue-lightness-chroma space
YCbCr.

Figure 9.8 summarizes the conceptual relationship between relative

brightness (B) and saturation (S) to lightness and chroma of a surface,
for a single-hue triangle in a hue-chroma- lightness space. The colours
F to Z show a range of intermediate tints between "Monitor Red" and
White. All have maximum relative brightness (100), because all are the
brightest possible version of that colour at that saturation. The
measured lightness (L) however ranges from 54 for pure red to 100 (by
definition) for white. The colours B to F have R = 51, 102, 153, 204 and
255 respectively; G and B = 0 in all cases. As light, these colours all
have the same (maximum) saturation, because they are all pure red; in
each case, only the red phosphors are glowing. As surfaces, they are not
all equal in chroma however - F is the strongest colour, the one most
different from grey.
Figure 9.8. Left: Conceptual relationship of lightness, absolute brightness,
relative brightness (B), chroma and saturation. Letters refer to colour series
discussed below. Right all the possible variants of a single hue angle (H).

Figure 9.9 Intermediate tints between pure red and white. The colours are all
at maximum relative brightness (B=100), because in each case they are the
brightest possible version of red at that saturation. They progressively increase
in lightness however from left to right.
Figure 9.9. Intermediate shades between black and pure red. The colours are
all at maximum saturation (S=100), because all are pure red. They
progressively increase in relative brightnessand chroma however from left to
right.

Uniform saturation series such as B to F are of interest because within

such a series, the brightness changes but the ratio of R to G to B
components, and hence the balance of wavelengths, does not change.
They therefore contain the set of colours that we need to represent a
surface of a single colour under different amounts of illumination.
Figure 9.11. Four examples of uniform saturation (S) series.

<< 1 2 3 4 >>

BRIGHTNESS-BASED COLOUR SPACES 3: HLS (=HSL) AND

HSI

Figure 9.12. HLS and HSI colour spaces. (A) HLS and (B,C) two views of HSI
colour space.Both spaces could be represented with either a circular or a
hexagonal cross section: the crucial difference between HLS and HSI is in the
different levels assigned to the pure colours, resulting from different definitions
of "L" and "I".

HLS and HSI are two other colour spaces encountered in graphics
applications (Figure 9.12).The parameter L in HLS space has a
particularly tenuous connection with perceived lightness. It is given by
the formula (maximum of r,g,b - minimum of r,g,b)/2, which results in
all fully saturated colours, irrespective of how light or dark they look,
having an L of 0.5. So-called saturation (S) in HLS is also calculated
very differently from S in HSB, and is essentially the degree of
saturation compared to the maximum possible at a given value of L.
Thus for example a very pale pink can have an S of 100. HLS is the
colour space used in the desaturate command in Photoshop, which
reduces all colours to a grey of the same "L" in HLS, and thus has an
entirely different effect to converting to greyscale mode, which (with
far more realism tonally) converts to a grey of the same CIE
lightness.HLS is the basis of the colour picker in Corel Painter (despite
the confusing labelling of the dimensions as H,S and V!).

The pseudo-lightness dimension I in HSI is given by the formula

(r+g+b)/3, which results in Monitor Red, Green and Blue having an I
of 0.33, and Monitor Yellow, Magenta and Cyan having an I of 0.67.
Saturation (S) in HSI is calculated in yet another way from S in HSV
and HLS.

<< 1 2 3 4 >>

Next: Part 10: Principles of Colour

Part 10. PRINCIPLES OF COLOUR FOR

PAINTERS
 1. Shading series
 2. Consistency of relative brightness
 3. The scale of "brilliance"
 4. Effects of coloured illumination
 5. Effects of multiple light sources
 6. Effect of distance from light source
 7. Effect of inclination to light
 8. Effects of atmosphere
 Applying the principles in paint

The dimensions of saturation and relative brightness turn out to be

vitally important when it comes to understanding colour and light, and
when painting effects of light from the imagination. In the final section
of this work I will summarize what seem to be the main considerations
under eight major principles. Many of these principles have been set
out somewhere or other before, most notably I believe in a series of
papers by Arthur Pope (1922, 1931), but they deserve to be a lot better
known than they are.

Edit, July, 2015: For an inexpensive and easy-to use app that models
the effects on shading series of any combination of lighting and object
colours, and outputs the results as swatches that can be imported into
digital painting programs, please check out Murray
Lancashire's Colour Constructor.

1. SHADING SERIES

Uniformly coloured surfaces passing between different levels of

illumination
under a single light source are represented by image colours that
move along a
line of uniform saturation, such that chroma increases as the value
increases.

If a surface of uniform colour moves or turns steadily away from a

single light source, it receives progressively less light per unit area, and
so the light diffusely reflected from it decreases correspondingly in
brightness. We would of course expect to paint this surface using a
series of image colours that decrease in value (lightness), but what
would happen to the chroma of these image colours?

Because the spectral make-up of the lighting and the spectral

reflectance of the surface are both constant, the light reflected from the
surface maintains the same relative proportion of wavelengths while
its brightness decreases. The colours of such a series of lights are said
to have the same chromaticity, which means that their hue and
saturation remain approximately the same. (The actual hue perceived
may shift slightly due to an effect called the Bezold-Brucke effect). To
represent this series in paint, we need a series of image colours, here
called a shading series, that follows this line of uniform saturation or
chromaticity. In such a series, chroma decreases steadily as value
decreases, at the precise rate needed to keep the saturation unchanged
(please see Fig. 9.6 !). Such series form straight-line paths radiating
from the point of zero light energy in simple colour spaces such as RGB
and YCbCr (Fig. 10.1.1B).

In L*a*b* space these paths generally drift in hue, and for much of
their length appear to radiate from a point 16 lightness units below
zero on the 100-unit lightness (L*) scale (Fig. 10.1.1C). In Munsell
space the paths both drift in hue and fluctuate somewhat in slope, but
overall they also appear to radiate from a point below Munsell value
zero (as first pointed out by Evans, 1974). I've speculated that this is
because Munsell value zero represents the light energy at the visual
threshold of blackness, while the chromaticity lines are radiating along
much of their length from the actual point of zero light energy (Briggs
in Flynt, 2010). Centore (2011) found that on fitting straight lines to
these paths in Munsell space, the average position of their apparent
point of origin was about one value step below zero on the Munsell
value axis.

Figure 10.1.1. A, Six uniform saturation series arranged in rectangles with a

greyscale surround, all organized to represent uniformly coloured fields under
varying illumination. David Briggs, 2007, Photoshop CS2. B, The uniform
saturation series from Fig. 10.1A plotted in YCbCr space, showing the pattern of
radiation from the zero energy point. C, The uniform saturation series from Fig.
10.1A plotted in L*a*b* space, showing the pattern of apparent radiation from a
point 16 lightness units below zero. Fig. 10.1.1B and C plotted using Colorspace.

It is remarkable that Figure 10.1.1A, which consists merely of an

arrangement of seven uniform saturation series (including the neutral
series), is sufficient to create a distinct visual impression of uniformly
coloured surfaces under varying illumination. To me, this diagram in
itself demonstrates the important role uniform saturation/chromaticity
series play in object-colour constancy. Evidently at a subconscious level
our visual system is attuned to recognizing these series within the
visual field, and presents them to us as perceived gradients of
illumination, though only when their spatial arrangement is consistent
with this interpretation (cf. Fig. 10.1.2). As I sometimes say to my
students: "You may not be interested in mathematics, but your visual
system is".

Figure 10.1.2. Another arrangement of the same colours as in Fig. 10.1A, this
time appearing as non-uniformly coloured shapes without any clear pattern of
illumination. David Briggs, 2014, Photoshop CS2.

In digital work we can easily create uniform saturation series

by keeping the Hue angle (H) and HSB Saturation (S) constant (in
effect keeping the ratio of R/G/B constant), while the HSB
brightness (B) decreases. Uniform saturation series are easily created
in Photoshop using the colour picker, which allows S and B to be
directly manipulated. In Photoshop shading series can also be created
by various shortcut methods such as

1. varying the opacity of a layer in normal mode over a layer of

black, or
2. varying the opacity of a layer of black in multiply mode over a
layer
3. by placing a coloured layer in multiply mode over a greyscale
image.
The first two methods work in RGB mode but not Lab mode.

In Figure 10.1.3 Principle 1 is used to draw a red sphere from the

imagination. The highlight retains the (white) colour of the light
source. Around the fringes of the highlight, additive mixing of this
white light with the red diffuse reflection creates intermediate colours,
while the darker versions of the red are restricted to the crevice
shadow.

Also note that, contrary to a widespread myth among painters (e.g.

Loomis, 1947), the richest colour is not on the edges of the lighted
area. In the case of Loomis this belief is connected with his belief that
the highlight coincides with the centre of the full light (link).
Figure 10.1.3. Colour relationships for a red ball on a white table. Specular
reflection on tabletop and sphere both move along lines of uniform saturation
between light and dark (Principle 1). Dotted lines shows table and ball
maintaining the same ratio of relative brightness in light and shadow (Principle
2). Sphere painted in Photoshop CS2.

According to this principle, B = 100 is the brightest version of any

colour that can be depicted- no colour in RGB space has a greater
brightness and the same balance of wavelengths. We will in fact see
that even painting a surface of full chroma, we will generally not use
colours as bright as B = 100 if we wish to leave room to represent the
highlight. The only way to go lighter than this is to try to give the effect
of a very bright light beyond the range of adaptation of the eye, and go
to paler, lighter colours, like those seen in an overexposed photograph.

In traditional paint mediums, the colour relationships discussed here

need to be established by eye. With practice it is not difficult to create
shading series in paint, and the inbuilt responsiveness of our visual
system to uniform saturation series, as suggested above, probably helps
here. Such series are not created by just adding black paint. Black
pigment tends to reduce the chroma of mixtures more rapidly than the
lightness, creating colours that are too neutral (grey) for their lightness
level - hence "traditional" colour theory's insistence on not using black,
because it "muddies" the colours. My advice is that anyone who doesn't
understand colour should not use black, but now that you've got this
far you're certainly ready to join the black-using party, who incidentally
form the majority of painters throughout art history. You understand
by now that painting is about creating colour relationships, not about
following colour recipes. In this case, the solution is to add just enough
of pure colour to get the mixture back onto a uniform saturation series
(Fig. 10.1.4). Some fine hue adjustment may also be needed to counter
any hue-shifting effects of the black, but the same is true whatever
pigment is used for darkening.
Figure 10.1.4. Adjustment of saturation with pure red pigment after adding
black pigment. Adding black to a mixture of Permanent Alizarin Crimson and
White draws the mixture away from the shading series into colours that are too
low in saturation. The solution is to add just enough extra Permanent Alizarin
Crimson to get the mixture back onto the line of uniform saturation. Some fine
tuning of the hue is also generally needed.

Incidentally, adding the complementary colour, "traditional" colour

theory's usual alternative to adding black for darkening, works
tolerably well for some colours for a small amount of darkening and
then starts drawing the mixture in the wrong direction; for other
colours it draws the mixture in the wrong direction right from the start
(Fig. 10.1.5).
Figure 10.1.5. Mixing of Cadmium Red and Cobalt Green. Cadmium Red and
Cobalt Green are near complementary in pigmentary mixing (B). Cadmium
Green initially draws Cadmium Red along a line of uniform saturation, then
begins drawing the mixture inward to less saturated colours; Cadmium Red
however draws Cobalt Green away from the required shading series (black
arrow) right from the start.

Revised July 16, 2014. Original (2007) text here

<< 1 2 3 4 5 6 7 8 9 >>

2. CONSISTENCY OF RELATIVE BRIGHTNESS

When several coloured surfaces pass together into different levels of

illumination they maintain constant ratios of relative brightness
(B).

If the brightness of appearance of two greyscale colours differs by a

factor of x at a given level of illumination, they will differ by the same
factor at any other level of that illumination (Figure 10.4). This is
because the difference in the amount of light energy reflected is
determined only by the relative reflectance of the two surfaces. This
rule works both with linear radiance and with nonlinear brightness (B).
It doesn't quite work with nonlinear lightness (L) because of the
different way L is calculated.
Figure 10.4. Greyscale colours in shadow and light, demonstrating
how relative brightness of appearance is maintained.

Note that the fact that relative differences of brightness are maintained
means that absolute differences between the different tones are less in
the shadows than in the light, and conversely that the absolute change
in tone between light and shadow is greatest for light tones and least
for dark tones. Note also that the frequently quoted rule that the
lightest tone in the shadow must be darker than the darkest tone in the
light is simply not true. This "rule" may possibly have arisen from a
simple misunderstanding of the principle that the lightest occurrence
of a given colour in the shadow will be darker than the darkest
occurrence of the same colour in the light.

In Photoshop, relationships of constant relative brightness can be

created very easily using either of the two shortcuts mentioned under
Principle One, varying the opacity of the image layer in normal mode
over a layer of black, or varying the opacity of a layer of black in
multiply mode over the image layer.

For coloured surfaces, the same rule of maintenance of relative

brightness also applies (provided that the light sources do not differ in
spectral composition). "Colorfulness" of appearance, and hence the
chroma of the image colour used to represent that appearance,
diminish with the diminishing brightness (Figure 10.5). The saturation
of the light given off by the surface, and of course
the perceived chroma of the surface itself remain the same.
Figure 10.5. Colour relationships in light and shade. Above. Effect of two
different light levels on appearance of a group of surface colours, and the same
colours in YCbCr, showing how patterns of relative brightness are maintained.
Below: Interactive demonstration of effect of different light levels on a set of
surface colours, emulated here by varying the opacity of layers over black.

<< 1 2 3 4 5 6 7 8 9 >>

3. THE SCALE OF BRILLIANCE

A coloured surface will look greyed, pure-coloured, fluorescent or self-

luminous depending on how its relative brightness (B) compares with
that of a white surface in the same lighting.
Figure 10.6 Demonstration of perception of "brilliance". Three yellow shapes
with relative brightness (B) of 53, 81 and 97, overlaid on photograph against
backgrounds at three different levels of illumination, where white paper is
represented by a grey of B = 53, 81 and 97 respectively. In each case the dot that
has the same relative brightness (B) as the surrounding paper looks like a
pure yellow, the less bright dots look olive-coloured, and the brighter dots look
fluorescent or luminous. Photograph by David Briggs.

Evans (1974) investigated how a surface, as it gives off progressively

more light compared to its surroundings, passes from looking greyish
to pure-coloured, then fluorescent and finally luminous. He suggested
the term brilliance for this scale and the term zero grey point for the
point where colours exhibit neither greyness nor fluorescence. I argue
below that this scale of brilliance has a direct relationship to relative
brightness (B) in an image, and that a colour is seen to have zero
greyness when it has the same value of B as what is perceived to be a
white surface under the same illumination.

We've already seen that white, all of the full chroma colours, and all of
the tints have a relative brightness (B) of 100. A consequence of
principle 2 above is that in a relatively shaded area, where we use a
grey with say, B = 50, to represent a white surface, all of the pure
colours and tints under the same illumination will also be represented
by colours with B=50 (Figure 10.6)
Figure 10.7. Aerial view of YCbCr space, showing the set of colours with B =
100 (left) and
B = 50 (right).

Putting this the other way around has a fascinating consequence. We

can see that in a setting where a white surface is represented by grey
with B of 50:

a. Any colour having B = 50 will read, depending on its saturation,

as either a full-chroma colour, a pure tint or white.
b. Any colour with B < 50 will look like a dark surface colour, i.e.
will exhibit a degree of greyness.
c. Any colour with B > 50 will look too bright to be simply reflecting
light. If it has only moderate excess brightness, such as a patch of
fluorescent paint might exhibit, it may have the appearance of a
fluorescent surface colour. However if it has a large amount of
excess brightness, it will read as being luminous, and will be seen
either as an independent light source, or as a specular reflection
of a light source.

<< 1 2 3 4 5 6 7 8 9 >>

4. EFFECTS OF COLOURED ILLUMINATION

The effects of coloured illumination follow the principles of subtractive

mixing, since only wavelengths present in the light source and not
absorbed by the surface can be reflected. Coloured lighting tends to
neutralize and darken complementary-coloured surfaces, thereby
tending to raise the relative lightness of all surfaces that reflect
wavelengths present in the light, the latter surfaces also shift towards
the hue of the coloured light (Figure 10.7). The range of colours seen is
always less than seen under white light. In Photoshop the effects of
coloured lighting can be suggested by an overlying coloured layer of
variable opacity in multiply mode; the greater the opacity, the stronger
the colour of the lighting.
Figure 10.8. Effects of coloured illumination. Left: Effects of green, red and blue
coloured illumination, at two intensities, on the appearance of various coloured
surfaces, simulated in Photoshop CS2 by an overlying coloured layer in multiply
mode. Right: Effect of strong green illumination, shown as displacements in
YCbCr space in side view (above) and in the CbCr plane. Colours lacking green
(purple and red) are drawn directly towards black; colours containing some
green drop in lightness, but less markedly, and shift in hue towards green;
colours fully saturated with green (yellow, cyan and white) all converge towards
bright green in hue and lightness.

Figure 10.9 shows the effect of coloured light on the colours of human
skin. Under white light (10.9B), human skin shows a range of low
chroma colours, often extending into slightly stronger colours in the
direction of red (where capillaries are numerous) and orange (where
pigment is denser). Incandescent light, being similar in hue to average
skin colour, shifts these hues shift to exhibit higher chroma but less
varied hue (10.9A). Under strongly bluish light, such as skylight, the
colours become more neutralized, but may exhibit a full range of hues,
including prominent crimson, greenish and bluish variants (10.9C).

Figure 10.9. Effect of different illuminants on human skin colours. Photograph

taken in white (flash) illumination (B), and transformed in ColorSpace to
simulate (A) warm incandescent illumination (illuminant A) and (C) cool
illumination (D75); resulting colours shown on CbCr plane of YCbCr space.

<< 1 2 3 4 5 6 7 8 9 >>

5. EFFECT OF MULTIPLE LIGHT SOURCES

Multiple light sources each create their own patterns of light and
shade. The quantities of light from each pattern combine additively,
but remember that it is light energy (linear radiance) that adds, not the
nonlinear brightnesses measured by B in Photoshop. An area lit by two
equal light sources gives off twice as much light energy as an area lit by
only one, but looks much less than twice as bright. One result of this is
that areas of overlap of shadows that get no light from either source
tend to be conspicuously dark compared to areas lit by one source
(Figure 10-10).
Figure 10.10. Effect of combination of two close point sources of light and weak
ambient illumination.

In Photoshop the effect of multiple light sources can be emulated

qualitatively by superimposing in screen mode layers representing the
light pattern created by each light source. This is how the images and
interactive animations demonstrating additive mixing on this site were
created (e.g Figures 10.10, 10.11).
Figure 10.11. Interactice animation created to demonstrate the effect of
multiple light sources (Figure 2.7). Sliders vary the intensity of one large white
overhead light source (topmost slider) and two small yellowish oblique light
sources. The sliders individually control the opacity of three layers in screen
mode, each painted with the light pattern generated by one light source. Each
layer may be viewed (against a black background) by moving the two other
sliders to the left-hand position. Animation created in Flash by Ray Kristanto
from a psd file painted in Photoshop CS2 by David Briggs. Copyright David
Briggs and Ray Kristanto, 2007.

The effect of mixing lights in screen mode often resembles additive

mixing qualitatively, though generally not quantitatively. The
nonlinear response of our visual system means that a light that has
twice the energy of another light will look brighter by a factor between
the cube root and the square root of two, i.e. around 1.37. When layers
are in screen mode, the difference between each RGB component and
one is multiplied by the corresponding difference in the other layer,
and the result subtracted from one. For example, if the two
components both have relative brightnesses of 0.5, the resulting
brightness will be 0.75, i.e brighter by a factor of 1.5. This factor
increases to nearly two for dim lights, and reduces progressively to one
(i.e. no increase) for bright lights. I assume that some sort of flattening
of response of this sort is inevitable, given that brightness has a finite
range in RGB space, unlike its open-ended range in the real world. In
any case, mixing of bright lights in screen mode can give quite different
results to additive mixing. For example, yellow and magenta at
maximum brightness mix to give pure white (Figure 10.12C), not the
whitish red that would result from additive mixing, and which you in
fact get if you mix them at 50% brightness (Figure 10.12B).

Figure 10.12. Attempted emulation of additive mixing of lights using layers in

screen mode, with lights at (A) 0.15, (B) 0.50 and (C) 1.00 of maximum
brightness. In A and B the resultant mixture is the expected whitish-red colour,
but is a little too bright to be accurate, especially for the mixture of dimmer
lights (A). The mixture of lights at maximum brightness gives white (C), which
is quite incorrect.

<< 1 2 3 4 5 6 7 8 9 >>

6. EFFECT OF DISTANCE FROM LIGHT SOURCE

The rate of decrease in the intensity of light with distance from a light
source depends on the size of the light source. All real situations lie
between two extremes:

1. Point source: Moving away from a point source of light, the

amount of light energy diminishes according to the square of the
distance (the inverse square law)
2. Infinitely large source: Moving away from a light source of
infinite extent (i.e. an infinitely large wall of light), the amount of
light energy does not change, irrespective of distance

In real situations this means that the fall-off of light energy is close to
an inverse square relationship for small light sources, and less rapid for
very large light sources. This fall-off applies to linear (light energy)
units, so you need to convert to nonlinear (perceived brightness) units
if this is the kind of unit you are using. The table below uses nonlinear
conversion used to calculate the nonlinear units in which RGB
"brightnesses" are expressed in graphics programs such as Photoshop.

Relative
1 2 3 4 5 6 7 8 9 10
distance
Light energy
100 25.00 11.11 6.25 4.00 2.78 2.04 1.56 1.23 1.00
(%)
Brightness
100 54 37 27 23 20 17 15 14 13
(%)

Table 10.1. Relative fall-off of radiance and brightness with distance from a
point source of light.

Figure 10.13. Fall off of brightness with distance, calculated using the
proportional reduction of brightness with distance according to the inverse
square law given in Table 10.1.

<< 1 2 3 4 5 6 7 8 9 >>

7. EFFECT OF INCLINATION TO LIGHT

A plane directly facing a light source receives the maximum flux of
light; but as the plane rotates, the amount of light striking a unit area,
and hence the amount of light energy reflected, diminishes in
proportion to the cosine of the angle of rotation. For example, a plane
inclined at 60 degrees to the direction of light fall catches half the light
energy that a plane facing the light source catches. As in the previous
section it is the linear (physical) light energy that is diminished by this
factor, so once again you need to convert from this if you are working
in nonlinear (perceived brightness) units, such as the RGB units seen in
Photoshop. Table 10.2 gives the fall off for the simplest possible
situation, where the light is an infinitely distant point source. For other
situatuions the light may terminate at angles greater or less than 90
degrees, depending on the size and distance of the light source. If
ambient illumination is to be considered, it should be added to the
linear energy values before conversion. For an object very close to a
light source, the fall-off may include a significant contribution from
increasing distance from the light source in addition to inclination.

Inclination
0 10 20 30 40 50 60 70 80 90
(degrees):
Light
energy 100 98 94 87 77 64 50 34 17 0
(%)
Brightness
100 99 97 94 89 82 73 62 45 0
(%)

Table 10.2. Relative fall-off of radiance and brightness of reflected light with
angle of inclination to direction of light.

For practical painting purposes this sort of calculation may not be

necessary unless particular accuracy is needed (all of the sphere
illustrations on this site, for example, were modelled purely by
"eyeballing").

This fall off of brightness with inclination to the light source is of

course the basis of our efforts to model form in drawing and painting
using tone. An interesting point to note in the table is how slowly
apparent brightness diminishes at low to moderate inclinations away
from the light, which is why such a large area can be treated, at least as
a first approximation, as a simple "full-light" zone.

In digital painting, the problem of synchronizing the change in

brightness of the different components of a multicoloured surface can
generally be solved very simply if the way the colouration behaves is
analyzed and understood. In Figure 10.14, an imaginary strip of apple
skin was conceived as having an underlying uniform green colouration,
modified in patches by varying concentrations of red pigmentation.
The green component was therefore modelled (using the principle of
uniform saturation) in one layer (Figure 10.14C), and the red
component was painted in an irregular pattern of overlapping
brushstrokes on an overlying layer in multiply mode (Figure 10.14B).
Figure 10.14. Painting a complex multicoloured surface pattern turning out of
a light source in Photoshop.

<< 1 2 3 4 5 6 7 8 9 >>

8. EFFECTS OF ATMOSPHERE

If objects are located in a turbid medium, such as a fog or murky water,

their hue, chroma and lightness all converge on those of the medium
with increasing distance from the viewer. In Photoshop this effect can
be emulated very simply by either interposing layers at low opacity
between the successively more distant objects (Figure 10.15), or by
achieving the same result with a single layer modified by a mask.

This basic principle, like many here, was clearly set out be Arthur Pope.
Pope also noted that the situation is more complex for atmospheric
perspective, where the blue colour is derived by scattering from the
light passing through the medium. In that case, for light toned objects
the addition of bluish light from the atmosphere is often exceeded by
the removal of bluish light by scattering on the way to the eye. Thus
light toned objects tend to become somewhat warm-hued, at least up to
the middle distance (Pope, 1931). This effect can be emulated in
Photoshop by using an orange layer in multiply mode, masked to
respond to depth and lightness (link)
Figure 10.15. Effect of coloured fog. Left: emulated in Photoshop CS2 using
interposed layers in normal mode at low opacity. Right: Colors from these
spheres viewed from two directions in YCbCr space.

<< 1 2 3 4 5 6 7 8 9 >>

APPLYING THE PRINCIPLES IN PAINT

These eight principles apply to painting in any medium where creating

a sense of light is important, but they are easiest by far to put into
practice in a programme like Photoshop, where parameters
representing hue, saturation, lightness and relative brightness can be
directly manipulated. They are far less easy to put into practice in
traditional mediums, where these parameters must be manipulated
indirectly by physically mixing coloured poisons, but this doesn't make
them any less important to any tonal painter. To create effects of light
from the imagination in paint, an option of course would be to first
establish colour relationships in a sketch version in Photoshop, and to
print this file as a guide to the colours needed.
I will not attempt, or presume, here to go on to present a complete
guide to painting, although some of the basic effects to watch for in
mixing paints have already been considered (link). (This of course is
the point where I tell you - as the saying goes - that you would need to
do one of my workshops in order for me to guide you through all of the
practical issues involved). Instead I will content myself for now to
conclude with some very brief remarks on some of the basic strategies
used for organizing a palette that painters use to make colour mixing
more efficient.

The most basic approach is simply to lay out colours straight from the
tube, and to mix each required colour individually. Hopefully this
mixing will involve visualizing the likely effect on hue, chroma and
lightness of each pigment before it is added, and systematically guiding
each colour through colour space to its target. There is nothing
inherently wrong with this approach except for inefficiency and
slowness.

A more efficient approach is to systematically pre-mix "strings" of

colour mixtures in order to have available an array of pools of colours
that can be used to draw a mixture through colour space in any
required direction. A simple and popular approach is to add white
progressively to each pigment (Figure 10.16). A more methodical
version of this approach involves controlling the precise lightness of
each of these steps, often in reference to the Munsell tonal scale.

Figure 10.16. Simple example of a palette arrangement created by adding

white progressively to black and several coloured pigments. Any required
colour can by mixed from one or two coloured pools, adjusted for lightness and
chroma using a grey pool.

My own preference is a variation on the set palette idea in which

colours are mixed along shading series (Figure 10.17). This makes the
mixing of the shading progressions needed in a tonal painting a lot
more efficient. Intermediate colours can be cross-mixed from these
pools, and additional shading series can be added for important
coloured surfaces in the picture if needed. Mixing such a palette takes
more time than the adding-white type, but once mixed the palette
permits very fast and fresh painting.

Figure 10.17. Simple example of palette arrangement created by

mixing shading series from several pigments. Any required colour can by
cross-mixed from one or two coloured pools and neutralized as needed using a
grey pool.

<< 1 2 3 4 5 6 7 8 9 >>

Next:References
REFERENCES
Albers, J., 1975. Interaction of Color: Revised Edition. Yale University
Press.

Arend, W.E., 1993. Mesopic Lightness, Brightness and Brightness

Contrast. Perception & Psychophysics, 54 (4), 469-476.

Aristides, J., 2007. Classical Drawing Atelier: a Contemporary Guide

to
Traditional Studio Practice. Watson-Guptill.

Benson, W., 1868. Principles of the Science of Colour. Chapman and Hall,
London.

Billmeyer, F.W. Jr. and Bencuya, A.K., 1987. Interrelation of the

Natural Color System and the Munsell Color Order System. Color
Research and Application, 12 (4) 173-186.

Birch, T., 1757. The History of the Royal Society of London, vol. 3. Millar,
London.

Blakeslee, B. & McCourt, M.E., 2015. What visual illusions tell us about
underlying neural mechanisms and observer strategies for tackling the
inverse problem of achromatic perception. Front. Hum. Neurosci., 21
April
2015. http://journal.frontiersin.org/article/10.3389/fnhum.2015.0020
5/full

Boyle, R., 1664. Experiments and Considerations Touching Colours.

http://www.gutenberg.org/etext/14504

Brewster, D., 1831. Treatise on Optics. Longman, Rees, Orme, Brown and
Green, London.

Brucke, 1866. Die Physiologie der Farben. Leipzig.

Carpenter, H.B., 1915. Colour. B.T. Batsford, London.

Centore, P., 2011. Shadow Series in the Munsell System. Color Research
and Application, 38(1), 58-64.

Cleland, T.M., 1921. A Practical Description of the Munsell Color System,

with Suggestions for its Use. Munsell Color Company,
Boston. http://www.applepainter.com/ (online edited version)
Evans, R., 1974. The Perception of Color. Wiley-Interscience, New York.

Fairchild, M.. D., 2004. Color Appearance Models: CIECAM and

beyond.
IS&T/SID 12th Color Imaging
Conference. http://www.cis.rit.edu/fairchild/PDFs/AppearanceLec.pd
f

Fairchild, M. D., 2005. Color Appearance Models (2nd Edn). Wiley

IS&T.

Field, G., 1817. Chromatics, or, an essay on the analogy and harmony of
colours [1st edn].

Field, G., 1835. Chromatography; or, a Treatise on Colours and Pigments .

Moyes and Barclay, London.
http://books.google.com/books?id=PDMAAAAAQAAJ. (1841 edn).

Ford, A., and Roberts, A., 1998. Colour Space

Conversions. http://www.poynton.com/PDFs/coloureq.pdf

Foss, C.E., Nickerson, D. and Granville, W.C., 1944. Analysis of the

Ostwald Color System. Journal of the Optical Society of America, 34(7),
361-381.

Gage, J., 1993. Colour and Culture. Practice and Meaning from Antiquity to
Abstraction. Thames and Hudson.

Gage, J., 1999. Colour and Meaning. Art, Science and Symbolism. Thames
and Hudson.

Goethe, J. W. von, 1810. Zur Farbenlehre (tr. Charles Lock Eastlake,

1840, as Theory of Colours).

Green-Armytage, P., 2006. The Value of Knowledge for Colour

Design. Color Research and Application, 31 (4), 253-269.

Gurney, J., 2010. Color and Light, a Guide for the Realist Painter. Andrews
McMeel Publishing.

Hanrahan, P. and Kruger, W., 1993. Reflection from Layered Surfaces

due to Subsurface Scattering. Princeton University, Department of
Computer Science,
Technical Report 409-93.
Hardin, C. L., 2014. More Color Science for Philosophers, in Dustin
Stokes, Mohan Matthen,
and Stephen Biggs (eds) Perception and Its Modalities. New York: Oxford
University Press. 379-389.

Hardy A. C., and Wurzburg F. L., 1937. Theory of three color

reproduction. Journal of the Optical Society of America, 27, p. 231.

Hatt, J.A.H., 1908. The Colorist (2nd edn, 1913). D.Van Nostrand
Company, New York.

Hayter, C., 1813. An Introduction to Perspective, Dialogues between the

Author's Children. Black, Parry and Co, London.
http://books.google.com/books?id=s-UHAAAAQAAJ (1815 edn)

Helmholtz, H., 1852a. On the Theory of Compound

Colours. Philosophical Magazine, Fourth Series, 4(4), pp. 519-34.

Helmholtz, H., 1852b. On Sir David Brewster's New Analysis of Solar

Light. Philosophical Magazine, Fourth Series, 4(27), pp. 401-416.

Helmholtz, H., 1866. Treatise on Physiological Optics. (English

translation by Optical Society of America, 1924–5).

Hering, E., 1878. Zur Lehre vom Lichtsinne. Vienna: Gerolds Sohn.
(English translation: Outlines of a Theory of the Light Sense, Leo
Hurvich and Dorothea Jameson, 1964, Harvard University Press).

Hurvich, L.M. and Jameson, D., 1957. An Opponent-process Theory of

Colour Vision. Psychol. Rev., 64, 384-404.

Itten, J., 1961. The Art of Color (1st English edn, tr. Ernst van Haagen).
Reinhold Publishing Company, New York.

Itten, J., 1973. The Art of Color (2nd English edn, tr. by Ernst van
Haagen). Van Nostrand Reinhold Company, New York.

Itten, J., 1970. The Elements of Colour (ed. by Faber Birren). Wiley, New
York.

Itten, J., 1975. Design and form: The basic course at the Bauhaus and later .
Wiley, New York.

Ives, H.E., 1935. The Ives Trichromatic Palette.

Kelly, K.L., 1943. Color Designations for Lights. Journal of Research of
the National Bureau of Standards. 31, 71-78.

Kemp., M., 1990. The Science of Art: Optical themes in western art from
Brunelleschi to Seurat. Yale.

Kerr, D.A., 2005. Color and Color

Spaces. http://dougkerr.net/Pumpkin/articles/Color_Models.pdf

Kirschmann, A., 1896. Color-Saturation and its Quantitative Relations.

American Journal of Psychology, 7, 386-404.

Kuehni, R. G., 2003. Color Space and its Divisions: Color Order from
Antiquity to the Present. Wiley-Interscience, New Jersey.

Kuehni, R. G., 2005. Color : an Introduction to Practice and Principles (2nd

edn). Wiley-Interscience, New Jersey.

Kuehni, R.G., 2010. Anonymous, Traité de la painture au pastel (Treatise on

pastel painting), in Anonymous (C. B.), Traité de la peinture en mignature
(Treatise on miniature painting), The Hague: van Dole, 1708. An English
translation with a speculative essay on its
authorship. www.iscc.org/pdf/TraitePastel.pdf

Kuehni, R.G., 2012. Unique Hues and Their Stimuli - State of the
Art. Color Research and Application, 39, 279-287.

Kuehni, R.G. and Schwarz, A., 2008. Color Ordered: A Survey of Color
Systems from Antiquity to the Present. Oxford University Press, USA.

Loomis, A., 1947. Creative Illustration. Viking Press, New York.

Loomis, A. 1951. Successful Drawing. Viking Press, New York.

Lowengard, S. 2006. The Creation of Colour in Eighteenth Century Europe.

Gutenberg-e, New York.
http://www.gutenberg-e.org/lowengard/

Luke, J. T., 2001. In The New Munsell Student Color Set, 2nd Edition by
Jim Long, Joy Turner Luke. Fairchild Books.

McLaren, K., 1985. Newton's indigo. Color Research and

Application, 10 (4), 225-229.

Marr, D., 1982. Vision, A Computational Investigation into the Human

Representation and Processing of Visual Information. MIT Press.
Mollon, J. D. 2003. The Origins of Modern Color Science. In Shevell, S.
(Ed) Color Science, Optical Society of America, Washington.

Moroney, N., Fairchild, M. D.; Hunt, R. W.G.; Li, C.; Luo, M. R.;
Newman, T., 2002. The CIECAM02 Color Appearance Model. IS&T/SID
Tenth Color Imaging Conference, Scottsdale, Arizona: The Society for
Imaging Science and Technology. ISBN 0-89208-241-0.
http://www.polybytes.com/misc/Meet_CIECAM02.pdf

Munsell, A. H., 1905. A Color Notation. An illustrated System Defining All

Colors and their Relations by Measured Scales of Hue, Value, and Chroma.
Boston.
http://www.google.com/books?id=PgcCAAAAYAAJ

Munsell, A. H., 1915. The Atlas of the Munsell Color System. Boston.

Newton, I., 1672. A Letter of Mr. Isaac Newton, Professor of the

Mathematicks in the University of Cambridge; Containing His New Theory
about Light and Colors: Sent by the Author to the Publisher from Cambridge,
Febr. 6. 1671/72; In Order to be Communicated to the R. Society .
Philosophical Transactions of the Royal Society, 6, 3075,8

Ostwald, W., 1916. Neue Forschungen zur Farbenlehre (New

Researches in Color Science) Physikalische Zeitschrift XVII, 1916, pp.
322-332, Leipzig: S. Hirzel (1899-1945). [Translated by Kuehni
in Kuehni, R.G. and Brill, M., 2010].

Ostwald, W., 1931, Colour Science. Part 1, Colour Theory and Standards of
Colour (translated by J. Scott Taylor). Winsor & Newton.

Palmer, G., 1777. Theory of Colours and Vision. Leacroft, London.

Palmer, S.E., 1999. Vision Science: Photons to Phenomenology. MIT Press.

Parkhurst, D., 1898. The Painter in Oil. Norwood Press, Massachusetts.

Parris, N.G.,1979. Adolf Hoelzel's Structural and Color Theory and its
relationship to the development of the Basic Course at the Bauhaus . Ph. D.
thesis, Uinversity of Pennsylvania, USA.

Poling, C.V., 1973. Color theories of the Bauhaus artists. Ph. D thesis,
Columbia University, USA.

Pope, A., 1922. Tone Relations in Painting. Harvard University Press,

Cambridge, Massachusetts.
Pope, A., 1931. The Painter’s Modes of Expression. Harvard University
Press, Cambridge, Massachusetts.

Pridmore, R., 2008. Chromatic Induction: Opponent Colour or

Complementary Colour Process. Color Research and Application, 33, 77-
81.

Purves, D. and Lotto, R.B., 2011. Why We See What We Do Redux, A

Wholly Empirical Theory of Vision. Sinauer.

Rood, O., 1879. Students’ Text-book of Color; or Modern Chromatics, with

Applications to Art and Industry. Appleton, New York, .
http://www.archive.org/details/Colour (1890 edn)

Ross, D. W., 1907. A Theory of Pure Design, Harmony, Balance, Rhythm.

Houghton, Mifflin, Boston
http://www.archive.org/details/theoryofpuredesi00rossuoft

Runge, P. O. 1810, Die Farben-Kugel, oder Construction des Verhaeltnisses

aller Farben zueinander. Hamburg: Perthes.

Ruskin, J., 1877. The Laws of Fesole. J. Wiley & sons, New York.

Schreiber, G., 1868. Die Farbenlehre. Für Architekten, Maler, Techniker und
Bauhandwerker, insbesondere für Bau- und polytechnische, höhere Gewerb-
und Realschulen, Leipzig.

Seim, T., 2013. CIE R1-57. The Border Between Blackish and Luminous
Colours. CIE Report 2013:R1-57.

Shapiro, A., 1984. The Optical Papers of Isaac Newton: Volume 1, The
Optical Lectures 1670-1672. Cambridge University Press.

Shapiro, A., 1994. Artists' Colors and Newton's Colors. Isis 85: pp. 600-
627.

Sherman, P.D., 1981. Colour Vision in the Nineteenth Century: the Young-
Helmholtz-Maxwell Theory. Adam Hilger Ltd, Bristol.

Froehlich, H.B. and Snow, B.E., 1904. Textbooks of Art Education, Book V,
Fifth Year. The Prang Educational Company, New York, Boston,
Chicago.

Spillman, W., 2009. Farb-Systeme 1611-2007. Schwabe Verlag Basel.

Wunsch, C. T. 1792. Versuch und Beobachtungen über die Farben des Lichts.
Leipzig: Breitkopf.

Young, T., 1802. On the Theory of Light and Colours, Philosophical

Transactions of the Royal Society of London 92, pp. 12-48.

Young, T., 1807. A Course of Lectures on Natural Philosophy and the

Mechanical Arts. J. Johnson, London.

CONTACT

About David Briggs

David Briggs obtained a Ph.D. in Science at the University of

Queensland in 1990. He has been teaching drawing and colour theory
at Sydney's Julian Ashton Art School since 1996 and has been a
lecturer and instructor at the National Art School, Sydney since 2009.
A page detailing his public courses is located here. Other significant
teaching positions have included six years as the life drawing instructor
for Disney in Australia (2000-2006), and lecturer in drawing at
Sydney's Billys Blue School of Graphic Arts (2006 - 2010). Since April
2015 he has been Chairperson of the NSW Division of the Colour
Society of Australia. A selection of his artwork may be viewed online
at here.

While I can be contacted by email, if you wish to ask me a question

about colour I would prefer you did so on my thread on the
forum conceptart.org, where the discussion might be of interest to
others.

E-mail

djcbriggs@gmail.com

Upcoming Courses
CLASSES AT MY STUDIO, CLOVELLY, NSW

 Colour, Light and Vision for NAS students (five-day

workshop)
 Sunday Landscape Painting
NATIONAL ART SCHOOL, SYDNEY

 Anatomy for Life Drawing (five-day workshop)

 Oil Painting with Colour and Light (eight-week evening
course)

JULIAN ASHTON ART SCHOOL, SYDNEY

 Colour, Light and Vision (five-day workshop)

 Fundamentals of Life Drawing (weekend workshop)
 Evening Life Painting and Drawing (ten-week
evening course)
 Beginning Life Drawing  (ten-week Saturday course)
 Essentials of Anatomy  (ten-week Saturday course)

ALL DETAILS:

https://sites.google.com/site/djcbriggs/classes

Links
General Colour Sites
http://djcbriggs.googlepages.com/generalcoloursites

Physics of Light and Colour

http://djcbriggs.googlepages.com/physicsoflightandcolour

Vision Science; Optical Illusions

http://djcbriggs.googlepages.com/visionscience

Colour Classifications, Specification, Spaces

http://djcbriggs.googlepages.com/colourclassifications,specification,spac

Colour and Vision Theory in Art

http://djcbriggs.googlepages.com/colourandvisiontheoryinart

Forums, Newsgroups
http://djcbriggs.googlepages.com/forums,newsgroups

Further Information
http://djcbriggs.googlepages.com/furtherinformation

MODERN COLOUR THEORY FOR

TRADITIONAL AND DIGITAL PAINTING
MEDIA
by Dr David. J.C. Briggs,
Julian Ashton Art School and National Art School, Sydney, Australia.
Chairperson, NSW Division, Colour Society of Australia.

 Site contents
 Acknowledgements

Colours of twenty two common artists pigments at various thicknesses

over a white ground. Photographed colours displayed in YCbCr space
using the program ColorSpace by Philippe Colantoni.

This website presents an account of the dimensions of colour and light

perception, written for painters using either traditional or digital media. The
conceptual framework presented here was developed as a component of Colour,
Light and Vision, a course in colour theory and practice for artists that I have been
presenting since 1998 at the Julian Ashton Art School, Sydney, of Theories of Colour, a
lecture course on the history of colour theory and practice that I presented at
the National Art School, Sydney, in 2009-2011, and of ongoing practical painting classes at both
schools.

Site Contents
All pages published in 2007 unless otherwise stated; latest revisions indicated in red:

Home

1 The Dimensions Introduced (revised Jan-Feb 2017)

 1.1 Colours in Space

 1.2 The Dimensions of What, Exactly? (added 15/1/17)
 1.3 The Dimensions of Colour: Lightness (added 16/1/17)
  1.4 The Dimensions of Colour: Hue (added 21/1/17)
 1.5 The Dimensions of Colour: Chroma (added 23/1/17)
 1.6 The Dimensions of Colour: Brightness and
Colourfulness (added 30/1/17)
 1.7 The Dimensions of Colour: Saturation (added 31/1/17)
 1.8 The Dimensions of Colour: Blackness and
Brilliance (added 15/2/17)

2 Basics of Light and Shade

 2.1 Specular and Diffuse Reflection

 2.2 The Zone of Light
 2.3 The Zone of Shadow

3 Basics of Colour Vision

 3.1 Introduction
 3.2 Trichomacy and Opponency (revised 08/10/11)
 3.2 Adaptation and Successive Contrast
 3.4 Colour Constancy
 3.5 Simultaneous Contrast and Assimilation
 3.6 What is Colour? (added 09/3/14)
 3.7 Answers to "What is Colour"? (added 09/3/14)

4 Additive Colour Mixing

 4.1 Additive Primaries (revised 5/8/12)

 4.2 Additive Mixtures (revised 5/8/12)
 4.3 Additive Complementaries (revised 5/8/12)
 4.4 Additive-Averaging Mixing (revised 5/8/12)
 4.5 Object Colours (revised 5/8/12)

5 Subtractive Colour Mixing

 5.1 Subtractive Mixing (revised 13/8/12)

 5.2 Ideal Subtractive Primaries (revised 12/8/12)
 5.3 Subtractive Complementaries (revised 26/6/12)

6 Colour Mixing in Paints

 6.1 Mixing Paints (revised 2/11/12)

 6.2 Primary Colours (revised 2/11/12)
 6.3 Paint-mixing Principles (revised 2/11/12)

7 Hue
  7.1 Hue from Aristotle to Newton (revised 15/4/13)
 7.2 The RYB Hue Circle or Artist's Colour Wheel (revised
15/4/13)
 7.3 Hue Systems Based on Opponent Colours (revised
28/7/13)
 7.4 Hue Systems Based on Additive Complementaries
 7.5 Hue Systems Based on Pigment-Mixing
Complementaries
 7.6 Orthogonal Systems
 7.7 Warm and Cool Hues

8 Lightness and Chroma

 8.1 Lightness
 8.2 Chroma
 8.3 Hue-Chroma-Lightness Colour Spaces

9 Brightness and Saturation

 9.1 Brightness, Saturation and Colorfulness

 9.2 RGB, CMY and CMYK Colour Space
 9.3 HSB (=HSV) Colour Space
 9.4 HLS (=HSL) and HSI Colour Space

10 Principles of Colour

 10.1 Shading Series (revised 26/7/14)

 10.2 Consistency of Relative Brightness
 10.3 The Scale of Brilliance
 10.4 Effect of Coloured Illumination
 10.5 Effect of Multiple Light Sources
 10.6 Effect of Distance From Light
 10.7 Effect of Inclination to Light
 10.8 Effects of Atmosphere
 10.9 Applying the Principles in Paint

11 References

12 Glossary (under construction)

Epilogue: Traditional and Modern Colour Theory (added

30/6/15
(Now also on
YouTube!: https://www.youtube.com/watch?v=4enFjTGVTnc )

 Modern Colour Theory (added 30/6/15)

  Traditional Colour Theory Strikes Back! (added. 30/6/15)

Acknowledgements
Warm thanks go to the "Dimensions of Colour team", Xavier Peria, Ray Kristanto, Noopur Patel,
Atania Trinata, and Debolina Bandyopadhyay from the 2007 second year Multimedia course at
the Billy Blue School of Graphic Arts, Sydney, and their teacher Dave Agius, for creating the site,
including the interactive animations. Thanks also to Ben Green for generously hosting the site
during its first year online, and to ibiblio, "the public's library and digital archive" at the
University of North Carolina - Chapel Hill, for accepting the site into their collection and hosting
it since then. Finally, thanks to all of those who have added links to this site on their websites,
blogs and forum posts, and especially to the following for their published comments:

Mark Fairchild (USA), Professor of Color Science & Imaging Science, Rochester Institute of
Technology, New York, and author of the textbook Color Appearance
Models (Wiley): Essentially an online textbook/tutorial on appearance, or "the dimensions of
colour and light" written from the perspective of artists. The site is very nicely done and blends
technical and artistic information well.

James Gurney (USA), illustrator, fine artist, author of numerous books including
the Dinotopia series and Color and Light: A Guide for the Realist Painter: David Briggs is none
other than the mastermind behind the website "Dimensions of Color." It's one of the best
resources on light and color on the Internet. I owe much of what I've learned on the topic to
[Dr] Briggs. ...His website "HueValueChroma" has that rare combination of depth and clarity.

David Gray (USA), fine artist and painting teacher: ... an absolutely indispensable source of
color knowledge for the realist painter: HueValueChroma.com. All of you who have asked me
about color really need to visit this site and get this information into your artistic thought
processes. It's going to be a little rough going for some who shy away from technical language.
It's also going to challenge some of the conventional color "wisdom" that has been taught in
art schools for years. I personally find the information fascinating and VERY USEFUL ... I
hope HueValueChroma will give you more control over your color choices as it has me.

Douglas Flynt (USA), fine artist and painting teacher at the Grand Central Academy of Art,
New York:"Huevaluechroma.com" is a great resource to better understand color and how light
affects color.

Slade Wheeler (USA), fine artist. His site hosts a large amount of well organized/concise
information coupled with informative illustrations, including 3D modeling and animations, all
of which make this one of the best online color theory resources that I've been able to find.

Colour Research Society of Canada/Societe canadienne de recherche sur la

couleur (Canada). Excellent overview of the dimensions of colour and light perception for
painters & digital media artists;

Danny Pascale (USA), CEO of BabelColorR colour measurement and analysis: A well
illustrated site on light, color, and its perception. The content is a course in applied color
science optimized for artists but useful for all. The language is clear, with just a few simple
equations and lots of descriptions.

Mary-Angela Papalaskari (USA), lecturer, Department of Computing Sciences, Villanova

University, Pennsylvania. ... a set of webpages that give a great overview of color as it is
perceived, from the artist's perspective.

William Cromar (USA), artist, lecturer and Art Program Coordinator, Abington College, Penn
State University. "Color is a fascinating topic which we've only been able to scratch the surface
of in this title" [ART 314 - Material Culture: Light and Color]. "If you wish to go in greater
depth, visit David Briggs' comprehensive website The Dimensions of Colour."

Thomas Scholes (USA), digital artist and painting instructor, moderator

of Futurepoly forum: I think the best advice I can give is to study light in terms of physics, this
website is a great resource in those regards.

Josh Yavelberg, Professor, Art Institute of Washington (USA). The Dimensions of Color: an
in-depth guide on the concepts of the various color systems and how to move within each color
space by David Briggs. I urge you to go to the second, and onward, pages of each area as there
are interactive flash demonstrations of the various color spaces.

Chris Raadjes (UK), game artist at Auroch Digital Ltd. Probably the most intensely scientific
approach to light for painters [that's] available on the web. ... I'm struggling myself, but it's
worth it.

Daz Watford (UK), video game developer, concept artist: Now this website is big and
intimidating; but it's a great explanation of how colour created by light works. It's quite
sciencey and took me three goes to start to "get it", but it's worth the struggle. It will change
your understanding of colour with a "mind = blown" Inception ...

Paul Foxton (UK), fine artist, author of website Learning to See: ... this site has more
information than any site should really be allowed to have in one place. David's site is nothing
short of incredible. There's so much information there, and it bears such careful and close
reading, that I can only take it in bite sized chunks. I read half a page and have to think about
it for a week. This the best site about colour I know of. The relevance of all of it to painting may
not be apparent to you straight away, and it may appear too scientific for 'feeling' types. But I
find myself mulling over things I've read there as I work, and it always results in deeper
insights into the way we perceive light and colour. Very highly recommended.

ALISON online training (UK): This course is ideal for any learner who practices the visual
arts, either professionally or as a hobby, and who wants to greatly enhance their knowledge
and understanding of colour theory.

Atelier Art Classes, Brisbane (AUS): An incredible resource for the painter [and] a
fascinating and informative resource for anybody who has an interest in the perception of
colour.

Joe Collins, Draw Academy (USA):Â If you ever want to read up on [colors], David Briggs
gives the most complete treatment I've ever found, can't recommend that website enough.

Bjoern Gschwendtner, Classical Atelier @HOME (Germany): Read more about color
on huevaluechroma.com. This is THE RESOURCE for color in the internet for artists.

Michael Hosticka (USA), recent Game Art & Design graduate, Ringling College of Art and
Design: I learned more about practical application of color within 10 minutes of reading that
than I have in all of my art classes combined.... I would highly recommend the website to
anyone who wants to improve their understanding of light and color and doesn't mind
technical reading.

Unless otherwise indicated, all material on this website is copyright Dr David

Briggs, 2007-2015, and is licensed for personal and commercial use under the
terms of a Creative Commons Attribution-No Derivative Works 2.5 Australia license.
Interactive demonstration of additive colour mixing. Drag the top, middle and bottom
triangular yellow sliders to the left to control the brightnesses of the red, green
and blue spotlights respectively. Copyright David Briggs and Ray Kristanto,
2007.

Next: Colours in Space

PROLOGUE: TRADITIONAL AND MODERN

COLOUR THEORY
PART 1: MODERN COLOUR THEORY
Figure 1.1. Extracts representative of traditional colour theory, from Johannes
Itten's The Art Of Color (1961, pp. 34, 78 and 22).

Colour training in the arts today is curiously divided between

traditional and modern colour theory. Traditional colour theory
characteristically begins with the concept of three primary
colours identified as red, yellow and blue.
These historical primaries rose to prominence over the course of the
17th century, and then as now were based on the observation that
mixtures of a wide range of colours can be obtained from colourants
(paints and dyes) of just a reddish, a yellowish and a bluish
colour, along with the psychological perception that red, yellow and
blue seem pure or "simple" compared to most other colours. A second
characteristic of traditional colour theory in its typical form is that it
speaks as if colours themselves intermix when we mix colourants, so
that the colour of a paint mixture is made of the colours of its
component paints. From this viewpoint each of the three
"secondary" colours made by mixing colourants of two primary colours
is made of those two colours, for example the colour orange is
physically made of the colours red and yellow and the colour green is
made of the colours yellow and blue. Black, grey and greyish colours
are commonly held to contain all three primaries, since they can be
made by mixing a secondary with the remaining primary, called its
opposite or complementary colour (Fig. 1.1). Greyish colours are
sometimes called "tertiary" to reflect this belief that they are made
of threecomponents, although this term is also used in traditional
colour theory in a different sense, for the six third-order colours that
lie between adjacent primaries and secondaries, and so supposedly
"contain" two primaries in unequal amounts. By extrapolating from
these assumptions, red, yellow and blue, along with white and
sometimes black, are deemed to be the fundamental components of all
colours, and paints of these three colours should "in theory" yield all
possible colours by intermixture. The fact that they do not can be
blamed on a lack of "pure" primaries among available paints, and
clearly a bluish-red paint, which by this view contains the primary
colour red plus a visible "impurity" of blue, must be further from being
a good colourant-mixing primary than one that is perceptually pure
red.

Figure 1.2. A. Colour wheel from George Field's Chromatography (2nd edn,
1841). Numbers around the circumference indicate the proportions in which
secondary and tertiary colours must be present in order for the three primaries
to be balanced. B. Post-Newtonian model of traditional colour theory, in which
white light is thought to consist of red, yellow and blue rays. A mixture of yellow
and blue paints reflects only the yellow and blue rays, which are detected by
yellow and blue receptors in the eye, leading to a mixed sensation of green. C.
Three-dimensional colour space proposed in a 1758 lecture by Swedish
astronomer Tobias Mayer, in which colours are arranged in a double pyramid
according to proportions of red, yellow, blue, black and white components.

Traditional colour theory readily adopted Newton's hue circle of 1704

in the form of the artists' "colour wheel" with its symmetrically placed
historical primaries (e.g. Fig. 1.2A). More problematic was Newton's
conclusion that lights of all colours are composed of rays of the
innumerable but seven-named colours of the spectrum, and that green
and orange lights could be physically simple as well as compound. This
problem however could be countered by supposing that the spectrum is
not continuous, as Newton had thought, but is actually made of
separate red, yellow and blue rays that produce the remaining colours
by intermixture. The suggestion that our eyes have three kinds of
receptors responsible for colour vision was first made based on this
hypothesis of red, yellow and blue rays. In this enviably simple view of
colour, the historical colourant-mixing primaries coincide with the
primaries for mixing coloured lights, with the sensitivities of the visual
receptors, and with the psychological primaries of perception. For
example in Fig. 1.2B, a mixture of yellow and blue paints reflects only
the yellow and blue rays, which are detected by yellow and blue
receptors in the eye, leading to a mixed sensation of green. The three-
ray hypothesis was quite widely accepted in science, and the first three-
dimensional colour order systems for classifying object colours were
made on the assumption that all such colours are mixtures of red,
yellow, blue and white (+ black) components in various proportions
(Fig. 1.2 C). This hypothesis was not overturned until 1852, when the
version of it promoted by the Scottish physicist Sir David Brewster was
convincingly demolished by the brilliant German physicist and
physiologist, Hermann von Helmholtz.

In subsequent years Helmholtz, his contemporaries including James

Clerk Maxwell and Ewald Hering, and their successors collectively
transformed our understanding of what colour is and how it works as
fundamentally as Darwin transformed our understanding of biology
over the same period, and this new understanding in turn led to
modern colour printing, colour photography, cinema and television,
and eventually to digital photography, painting and rendering. Modern
colour theory, as understood on this website, is a system of guidance
resting on the legacy of this late 19th century revolution, adapted to the
practical needs of painters and teachers of painting.
Figure 1.3. The psychological, additive and subtractive primaries of modern
colour theory.

In line with Hering's widely accepted opponent model, modern colour

theory recognizes four psychological primaries, the three historical
primaries plus green, arranged in two opposing
pairs, yellow/blue and red/green (Fig. 1.3A). This model is founded
not on neuroscience but on contemplating how we experience colour:
only red, yellow, green and blue are experienced as pure hues, and all
other hue perceptions are combinations of adjacent pairs of these four:
yellow-red (orange), yellow-green, blue-green (cyan, turquoise, etc.),
and blue-red (purple, magenta etc.). Red, yellow, green and blue are
thus the true primary colours of modern colour theory, as opposed to
primary colour stimuli (of additive mixing), or primary colourant
hues (of subtractive mixing).
Figure 1.4. A. The four opponent hues or "psychological primaries", and the
range of their combinations evoked by single wavelengths of light, and therefore
seen in the spectrum. B. Origin of the colours of the spectrum from successive
combinations of red vs green and yellow vs blue colour-opponent signals
created in the brain.

These four opponent hues are not physical components of light, but are
perceptions created in the brain of the observer in the form of
yellow vs blue and red vs green colour-opponent signals. Blue is
evoked by the shorter and yellow by longer wavelengths of visible light,
while green is evoked by the middle wavelengths and red by the two
extremes (Fig. 1.4B). Successive combinations of these four hue
perceptions create the sequence of hues seen in the spectrum as well as
the nonspectral hues like magenta, and thus explain how we arrive at
this circular succession of hues that has no physical basis in
the linear sequence of wavelengths in the spectrum. The concept of
four psychological primaries has strongly influenced several important
colour order systems, notably the Ostwald system, which was the
standard for British education in the 1930's, and the Natural Colour
System (NCS), the current colour standard in Scandinavia and Spain,
and was foreshadowed in the writings of Leonardo, Goethe and others.

In addition to the four psychological primaries, modern colour theory

refers to three additive (or light-mixing) and three subtractive (or
colourant-mixing) "primary colours" for the colours of lights and
colourants respectively that yield a maximum gamut or range of
colours of their mixtures. None of the additive or subtractive primaries
bear a direct relationship to any of the psychological primaries, but we
have a strong unconscious predisposition to apply to the former
the names of these hues that form the framework of our experience of
colour. Thus before the twentieth century, when our best available
colourant-mixing primaries were slightly greenish yellow, bluish red
and greenish blue paints, we almost universally labelled these
primaries as simply "yellow", "red" and "blue". In the same way today
we routinely label the orangeish red, yellowish green and blue or violet-
blue primaries of additive-mixing technology as simply "red", "green"
and "blue" (RGB).

It is counterproductive to the wider acceptance of modern colour

theory, as well as misleading, to dismiss the red, yellow and blue
primaries as merely obsolete, or worse, to insist (by adopting an
idiosyncratic and unrealistic definition of the term) that primary
colours "do not exist". The three historical primaries were as much a
step towards recognition of the four modern psychological primaries as
they were an imprecisely-named set of colourant-mixing primaries.

Figure 1.5. A. Dominant wavelengths of sRGB "red", "green" and "blue" additive
primaries in relation to responses of the L, M and S cone cells. B. Mixing of RGB
additive primaries.

So while the number four is the key to colour perception, three is the
key to colour stimulus and colour technology. This number three
stems from the fact that colour vision is ultimately based on three types
of receptors in the eye called L, M and S cone cells. These cone cells are
often loosely termed "red", "green" and "blue", and were once believed
to "detect" the red, green and blue-violet bands of the spectrum, and to
directly create red, green and violet or blue "fundamental sensations"
respectively. However it is now known that the L, M and S cones do not
detect individual wavelength bands or "colours" of the spectrum, but
respond to very broad and extensively overlapping ranges of
wavelengths peaking in the parts of the spectrum we see as greenish-
yellow, green and blue-violet (Fig. 1.5A), and respond in exactly the
same way, just to different degrees, throughout their range. Nor do
they directly create sensations of individual colours.
Instead, differences in the responses of the three cone types are
recorded in the retina in the form of cone-opponent signals (L-M and
L+M-S). Colour as such is created in the brain in the form of colour-
opponent signals based indirectly on the wavelength-dominance
information contained in these cone-opponent signals. When a light
creates a balanced response of all three cone types, each cone-
opponent signal is balanced at zero, and the light is seen as colourless
("white light"). Colours of lights are perceptions created by our visual
system in response to an uneven distribution of wavelengths, as
detected by an unequal response of our three cone types.

RGB "red", "green" and "blue" lights work as additive (light-

mixing) primaries because each of them stimulates one cone
type more than the other two (Fig. 1.5A), and so when mixed in
different proportions they produce a large variety of relative L, M and S
cone responses, and therefore cone-opponent signals, and therefore
ultimately colours. "Red" and "blue" lights unsurprisingly yield
mixtures passing through magenta (red-blue), and "green" and "blue"
lights unsurprisingly yield mixtures passing through cyan (green-blue),
but, to the perennial astonishment of students, "red" and "green" light
mixtures pass through an apparently new colour, yellow, that
is neither reddish nor greenish (Fig. 1.5B). This asymmetry is the
inevitable consequence of the mismatch between having three additive
and four psychological primaries. The yellow is not really a
“new― colour, in that it is already present in the colours evoked
by the component lights (they are orange-red and yellowish green
after all!), and when these are mixed the yellow signals add together
while the red and green signals cancel out, leaving pure yellow. The
yellow colour is not made of the colours red and green, and is
certainly not some sort of mistaken guess that the light has a
wavelength that lies between those of red and green.
Figure 1.6. A. Subtractive mixing of digital subtractive primaries. B. Explanation
of subtractive mixing of digital subtractive primaries.

Helmholtz recognized that the "colour mixing" of two colourants was in

large part a subtractive process in which the colour of the mixture
depends on which wavelengths both colourants pass on. It was soon
realized that in theory three transparent colourants that each
completely absorb one additive primary and completely pass on the
other two would, when mixed subtractively in varying proportions,
yield the entire gamut of those additive primaries, and hence be
optimal for subtractive mixing. These ideal subtractive primaries would
therefore resemble the pure yellow ("red" plus
"green"), magenta ("red" plus "blue") and cyan ("blue" plus "green")
colours of computer screens (Fig. 1.6), the last two being far from the
red and blue primaries of traditional colour theory. It took many
decades however before permanent pigments at all close to ideal
magenta and cyan were developed, and even today our best magenta
and cyan paints and inks are redder and bluer respectively than ideal
subtractive magenta and cyan. (Fig. 1.7A).
Figure 1.7. A. Mixing paths of a yellow, a magenta and two cyan colourants,
calculated using the program drop2color by Zsolt Kovacs Vajna. (For mixtures
using a greener yellow see Fig. 2.7, next page). B-D. Spectral reflectances of
pairs of cyan, magenta and yellow colourants and representative mixtures,
calculated using the same program.

Figure 1.7 shows predicted mixing paths and reflectance profiles of

some cyan. magenta and yellow paints, calculated and plotted using the
program drop2color by Zsolt Kovacs Vajna. Consistent with modern
colour theory, the paints produce a larger and more evenly distributed
gamut (range of colours) than the historical red, yellow and blue
primaries (see Fig. 2.6, next page). Note well that in subtractive
mixing, yellow and cyan colourants make green not because green
"contains" yellow and cyan, as typical traditional colour theory
assumes, but because both yellow and cyan colourants pass on large
amounts of "green" wavelengths (Fig. 1.7B).
Figure 1.8. A. Additive complementary pairs: digital "red" and "cyan", digital
"green" and "magenta" and digital "blue" and "yellow". B. Comparison of
opponent pairs and additive complementaries. In this CIE xyY (chromaticity)
diagram, additive complementaries are connected by a straight line through the
central "white point", in this case Illuminant C (after Kelly, 1943). Middle yellow
and middle blue are essentially complementary, but middle red and middle
green are not.

Just as different sets of colours are considered primary for perception,

for mixing lights and for mixing colourants, modern colour theory also
recognizes somewhat different pairs of "opposite" colours in these
three contexts. For every imbalance of cone responses that we
experience as a colour of light, there are lights that create an opposite
or complementary imbalance, and which therefore by mixing in the
right proportion can create white light. Examples of such pairs of lights
include the digital red-cyan, green-magenta and blue-
yellow pairs familiar from graphics programs (Fig. 1.8A). Perhaps
because the relationship between the cone-opponent responses and the
colour-opponent signals we ultimately experience is indirect, pairs of
colours of lights that mix to make white light (additive
complementaries) do not always coincide with the axes of our
perceptual framework of opponent colours determined by
experimental studies of large numbers of observers. According to such
studies the yellow - blue opponent pair is also additively
complementary or nearly so, but the red - green opponent pair is not
(Fig. 1.8B).

Figure 1.9. A-D, Mixing of subtractive complementaries (Liquitex Acrylics). E-G

Mixing paths of a cyan paint, Light Blue Permanent (E), a magenta
paint, Medium Magenta (F) and a yellow paint, Cadmium Yellow Medium
Hue (G), with other paints in the collection. All diagrams produced using the
program drop2color by Zsolt Kovacs Vajna.

Colours of pairs of colourants that make neutral (black or grey)

mixtures are called subtractive or colourant-mixing complementaries.
The exact result of subtractive mixing depends on the precise
wavelength-by-wavelength absorption of the colourants, and can not
be exactly predicted from their colours alone, but the importance of
this proviso is sometimes exaggerated as far as actual colourants go,
and it does not prevent reasonably reliable generalizations being made
about colourant-mixing complementaries. Subtractive and additive
complementaries coincide quite closely near the red-cyan and green-
magenta axes, but differ substantially near the yellow-blue axis (Fig.
1.9A, E). A yellow and a blue paint that are opposite on the Munsell hue
circle would be approximate additive complementaries, so that when
lights from these paints are mixed (for example, using spinning discs)
they would make a near neutral (see Fig. 4.4.1), but such paints
nevertheless reflect enough green wavelengths in common to make a
mid-chroma green mixture when combined physically. It is this
generation of green as an apparent compound of two visually distinct
colours that has influenced painters since Leonardo to reject or
question green as a "simple" or primary colour.

To make a mixing complementary for a yellow paint generally requires

a precise mixture of blue and violet paints (as in Fig. 1.9A) that is more
reddish than the blue additive complementary, but less reddish than
the middle purple expected in traditional colour theory (e.g. Fig. 2.1,
next page). Interestingly, afterimage complementaries show a
similar sort of pattern to these paint-mixing complementaries, in that
they agree well with the additive complementaries near the red-cyan
and green-magenta axes, but depart in a similar way near the yellow-
blue axis. Complementary-coloured afterimages result from changes in
the relative adaptation of the three cone types, and act as coloured
virtual filters, subtractively modifying the areas of the visual field
they occupy.
Figure 1.10. A,B,C, External view, hue plane and value-chroma plane of the
Munsell system. D,E, side and top view of colour solid devised by Arthur Pope
based on the colour classification of his former teacher, Denman Ross. F,G,
Ross' subdivision of the colours on a hue page according to chroma relative to
the maximum possible ("intensity") and chroma relative to the maximum
possible at that value ("neutralization"). H. Pope's subdivision of a hue page
according to "purity" (later called saturation) and "energy of vibration" (later
called brightness and then brilliance). I,J, External form and cross section of
the Ostwald system, showing divisions according to proportional black, white
and colour content.

Modern colour theory, like modern colour science, makes constant

practical use of concepts of three-dimensional colour space. The
framework of hue, value (lightness) and chroma (colour strength)
devised by the American artist and art teacher Albert Munsell (1858-
1918) for classifying object colours has proved especially useful to
painters. Munsell explained his system in his small book A Color
Notation in 1905, and embodied it as a published Atlas of colour
samples in 1915. The Atlas was succeeded by the Munsell Book of
Color (published from 1929 to the present), and following extensive
testing and refinement by the Optical Society of America in 1943, has
become a very widely accepted standard in science and industry. The
Munsell Book of Color, or any of several less expensive alternatives,
enables a painter to examine almost every colour obtainable from paint
mixtures set out in a logical, value-based arrangement. On this site I
show how the system makes a potent framework for observing colours,
for understanding colour mixing in paints, and for creating effects of
light from the imagination. While painters most commonly use only
the conceptual framework of relative hue, value and chroma rather
than the absolute Munsell dimensions, there are strong practical
benefits to at least using an absolute scale of value like the Munsell
scale.

A simpler system of hue, value and relative chroma devised by

Munsell's friend and rival Denman Ross has also been widely used in
art education, and was further developed by Ross' former student
Arthur Pope (1921). Pope showed how hue pages could also be divided
up according to the saturation ("purity") and brightness ("energy") of
the reflected light producing each object colour (Fig. 1.10 D-H), in a
similar way to that later adopted in the HSB (or HSV) colour space of
graphics programs. Lines of uniform saturation are of great importance
to the painter in that they correspond to shading series, or the paths of
the image colours needed to represent an object of one colour under
different levels of illumination. Colour spaces such as the historical
Ostwald system (Fig. 1.10 I,J) and the modern Natural Colour System
(NCS) that lack a dimension of greyscale value are less useful for most
painters, but have found favour in many areas of design.
Figure 1.11. Digital colour spaces A. RGB: relative brightness scales of the
"red" (R), "green" (G) and "blue" (B) component lights. B. Ideal CMY. C. HLS:
simple double cone model using digital hue (H), a pseudo-value
dimension (L) and relative chroma (S). D. HSB (also known as HSV): digital
hue angle (H), and measures of saturation (S) and brightness relative to the
maximum possible (B). E. YCbCr: used for jpg compression and as a space for
some operations in Photoshop. F, G. Lab: perceptual space with dimensions of
greyscale value (L), reddish/greenish chroma (a/-a) and yellow/blue
chroma (b/-b). Lab space can also be expressed as Lch: value (L),
chroma (C) and hue (H).

The series of colour spaces devised by the Commission Internationale

de l' Eclairage (CIE) are central to modern colour science and to the
development of all forms of digital colour technology including digital
painting. Most fundamental is the CIE XYZ space, which is the
foundation of colour management, and measures colours in relation to
three virtual "primaries" that are not colours, but are purely
mathematical transformations of actual lights. More commonly
encountered by digital painters are the CIE xyY or chromaticity
diagram (Fig. 1.7B), and CIE L*a*b* space, which arranges colours
similarly to the Munsell system, but describes their positions in terms
of reddish/greenish (a*/-a*) and yellowish/bluish (b*/-b*) chroma
instead of hue and total chroma. Photoshop uses a version of CIE
L*a*b* known as Lab space as an means for designating colours in the
colour picker, and as a working colour space. Digital painters can select
and apply modifications to their colours using a great variety of colour
spaces for different purposes (Fig. 1.11), though probably only a small
minority understand these spaces well enough to take full advantage of
this versatility. Many painters who use traditional media are familiar
with CIE colour spaces through the Handprint website by Bruce
MacEvoy, whose plots of watercolour paints as points on the CIE
L*a*b* and CIE CAM02 hue planes are quite widely quoted on painting
forums.

Figure 1.12. A,C. Diagrams by Arthur Pope (1921) conceptualizing the effect of
changing illumination (A) and of greyish atmospheric mist (C) on the
appearance of an array of uniformly coloured stripes. B,D. Realizations in
Photoshop of the relationships postulated in Pope's diagrams A and C
respectively.
Figure 1.13. A,B, Emulation of additive mixing of coloured lights (A) and effect
of translucent atmospheric mist (B) from the original upload of this site. David
Briggs, Photoshop CS2, 2007. C. Digital painting exercise in which students
examine a photograph to determine lighting and atmosphere, and then paint in
a simple object to be consistent with these. Class demonstration in Photoshop
CS2 by David Briggs, Understanding Digital Colour, Billy Blue College of
Design, 2008.

Principles governing the relationship of colour and light have

been an important part of artistic colour theory since the times of
Alberti and especially Leonardo, but these principles could be
expressed only verbally before modern models of colour space
provided a framework for their more precise formulation. In the early
1920's Arthur Pope set out a series of important principles of colour
appearance in terms of his own colour order system (Fig. 1.12), and
beginning in the 1930's Faber Birren used the framework of the
Ostwald and Munsell systems to describe colour relationships that
create various effects of iridescence, lighting and atmosphere. In the
middle decades of the century the American illustrator and highly
influential teacher Frank Riley devised an extremely elaborate system
for calculating effects of light and atmosphere within the framework of
the Munsell system. Applying Reilly's system in oil paint generally
involved the use of an extensive palette of "strings" of premixed colours
kept in tubes. Principles of this kind can now be applied very easily
using the controls available in digital painting programs (Fig. 1.12 B,D,
Fig. 1.13; for details see Section 10), and of course can be applied
automatically using rendering software.

Figure 1.14. Four popular late 19th-century textbooks that explained the
Helmholtz-Maxwell revolution in our understanding of colour for artists and art
students.

Numerous excellent texts explaining modern color theory for artists

began to appear soon after the Helmholtz-Maxwell revolution, and
those by Benson, Bezold, Rood, and Church in particular proved
popular enough to be reissued in multiple editions over several decades
(Fig. 1.14). One of the most influential was Modern Chromatics, or The
Students' Text-book of Color (1879) by American physics professor and
trained painter Ogden Rood, which is a revelation as to the level of
technical understanding of color and light that was considered
appropriate for academically trained artists of the time.
Figure 1.15. A. Colour wheel sketch from one of Degas' notebooks of the early
1880's, illustrated by Richard Kendall in Degas, Beyond Impressionism (1996,
p.101). Though mistaken by Kendall for "a rather inaccurate drawing of on
elementary colour wheel", the circle in fact shows additive complementary pairs
(vert opposite violet, jaune opposite indigo,
and orange and rouge opposite bleu) in an orientation that seems to link it
specifically to a diagram from the frontispiece of Ogden Rood's Modern
Chromatics (B).

It is difficult to say just how widely painters of this era used modern
colour theory, because most art historians in their research (and
teaching) seem to neglect fundamental elements of the artist's craft like
colour in the pursuit of more esoteric concerns. For example, author
Richard Kendall maintained that "there is no indication that [Degas]
joined those of his contemporaries who engaged in a more thorough
study of [colour] theory and application. On the contrary a previously
unpublished and rather inaccurate drawing of an elementary colour
wheel from one of Degas's notebooks of the early 1880's (fig. 104)
suggests a delayed interest in such matters" (Fig. 1.15A). The drawing is
not an inaccurate elementary colour wheel, but in fact shows the
recently established additive complementaries, probably transcribed
from Rood's Modern Chromatics whose French edition had been
published in 1881. Thus the painter Degas took a close interest in the
latest developments in modern colour theory, even though the art
historian and Degas specialist Richard Kendall did not.

These pioneering texts were followed over the course of the 20th
century by numerous others incorporating various aspects of our
developing understanding of colour, by Albert Munsell, Wilhelm
Ostwald, Arthur Pope, Faber Birren and many others. Ralph Evans' An
Introduction to Color (1948) and George Agoston's Color Theory and
its Application in Art and Design (1979, 1987) may be singled out
among the many that are still well worth reading today. The Munsell
system was disseminated throughout the 20th century in numerous
editions and reprintings of Munsell's A Color Notation and in its
successor, The New Munsell Student Color Set of 1994 by Joy Turner
Luke (unfortunately the text is less reliable in more recent editions),
through college textbooks such as Maitland Graves' The Art of Color
and Design (1941, 1951), and orally through teachers such as Frank
Reilly and his disciples. Although Reilly did not publish his Munsell-
based painting system himself, elements of it have been published by
several of his former students, most completely by Apollo
Dorian (1989). A simple introduction to many elements of modern
color theory is included in James Gurney's excellent Color and Light: A
Guide for the Realist Painter, which I recommend as a first step to any
students who have difficulty approaching my own site. For more
detailed accounts of colour science written for artists, and of historical
and modern colour order systems, the best and most reliable sources
are the books of Professor Rolf Kuehni, while Bruce MacEvoy's
enormous Handprint website provides online information and opinion
on a wealth of advanced topics.

Published June 30, 2015.

Next: Part 2: Traditional colour theory strikes back!

PART 2: TRADITIONAL COLOR THEORY STRIKES BACK!

 Forty eight traditional colour theory diagrams from a Google image
Figure 2.1.
search for "colour wheel", All forty eight would have been found to include some
highly inaccurate complementary pairs if they had been checked using the
simplest of tests, as in Fig. 2.3 A,B.
Although textbooks for academically trained artists ably explained
modern colour theory from its beginnings, traditional colour theory
persisted into the 20th century in other texts, especially those written
for amateur painters, and in the thinking of other artists, notably some
of the pioneers of abstraction (Kandinsky, Mondrian) and
expressionism. Throughout the first half of the century fully modern
and fully traditional colour theory texts coexisted, and many texts
presented both approaches. Since the 1960s however, traditional
colour theory has had a renewed and dominant influence on art
education, even at tertiary level, very largely in the form set down
in Johannes Itten's book The Art of Color (1961). Itten's book has been
so influential that it defines the limits of artistic colour theory for the
majority of sources on the internet today (Fig. 2.1): a symmetrical
traditional colour wheel, schemes of "objective" colour harmony,
dictionary meanings of colours, the limited subset of colour
phenomena known to early 19th century science (mainly successive
and simultaneous contrast), and very little on colour in relation to
representation of light and atmosphere. Traditional colour theory has
entered the digital realm in the form of apps like the widely used Adobe
Color CC (formerly Kuler), which calculates complementary and triadic
colour sets using a modified digital hue circle in which the hue angles
are intended to conform to the traditional colour wheel (Fig. 2.3B). As
a result of its half century of ascendancy, many artists today assume
that traditional colour theory has dominated art education
continuously since its origins, and assume modern colour theory is a
very recent intrusion.

Figure 2.2.
A. Colour star used by Itten as the basis of his teaching at the Bauhaus c.
1921, showing English translations of the 12 hue divisions he derived from his
teacher Adolf Hoelzel. Notice the symmetrical placement of yellow, cyan blue,
and a "Purple" (magenta) intermediate between "Crimson" and "Purple violet".
B. Traditional hue divisions adopted in The Art of Color.

Itten took many elements of his colour theory from that of his teacher
Aldof Hoelzel, but in his book of 1961 he left out almost all of the
elements that Hoelzel had derived from late 19th and early 20th
century science, and presented a version of traditional colour theory
and classification almost entirely fixated at an early 19th-century stage
of development. Thus although his "colour star" lithograph of 1921
(Fig. 2.2A), which he had used as the basis of his teaching of colour at
the Bauhaus (Itten, 1975, p. 33), follows Hoelzel's 12-hue system
derived from the scientist Wilhelm von Bezold, in The Art of Color he
reverted to a traditional system comprising evenly-spaced historical
primaries and secondaries, with six intermediates (Fig. 2.2B).

Figure 2.3.
A,B. Supposedly complementary red-green pairs from the first edition of
Itten's The Art of Color (A) and from the online Adobe Color CC (formerly Kuler)
colour circle (B). To see the true afterimage complementary of each colour,
focus on the central dot on the left for ten seconds and then immediately focus
on dot on the right. Most observers will see that the afterimage complementary
of the reds as more cyan, and the afterimage of the greens as more magenta,
than the supposed complement. C. CIE xyY plot of the primary red and
opposing secondary green colours from the 48 online colour wheels shown in
Fig. 2.1, using the program ColorSpace by Philippe Colantoni. Additive
complementaries fall on a straight line through the central white point. Both
sets show a broad scatter of chromaticities but are centred near the green and
red additive primary hues, placing them even further from being additive
complementaries than Itten's 1961 pair.

Itten's three primary colours are defined in his text as perceptually

pure hues, oddly enough in terms of the names of the four
psychological primaries of modern science, which otherwise escape
mention: "a red that is neither bluish nor yellowish; a yellow that is
neither greenish nor reddish; and a blue that is neither greenish nor
reddish" (emphasis mine; see Fig.1.1 on preceding page). Secondary
colours "contain" two primary colours and "must not lean towards"
either of their "components", and these secondaries are held to be
complementary to the remaining primary in paint mixing, in additive
mixing, in simultaneous contrast, as negative afterimages, and even in
expressive interpretation (Itten, 1961, p. 137). Thus according to the
logic of the system, primary red must be complementary in all of these
ways to a middle green that does not lean towards either yellow or blue.
In reality, as we saw previously, the additive and afterimage
complements of a middle (5R) red are bluish-green (about 5BG) rather
than middle green (5G), and a very similar relationship applies to
afterimage complements (Fig. 2.3 B). Yet accurate complementaries -
the colour that "the eye demands" after seeing another colour - are
accorded enormous significance in Itten's theory as the basis of what
he considered to be "objective" colour harmony, which in all its forms
depends on a balance of complementary opposites. Most online
examples of traditional colour wheels, including Adobe Color CC, place
the red-green pair close to the additive primaries called "red" and
"green", giving even more inaccurate additive and afterimage
complementaries (Fig. 2.3 C).

Itten does include two elements of modern colour theory on a single

page entitled "Color Physics" early in his book, but these are puzzling
to reconcile with the rest of his text. The two elements are the
continuous spectrum (described by Newton but in question until
upheld by Helmholtz against Brewster) and the idea that the mixing of
paints is "governed by the rules of subtraction". Itten's account of the
spectrum makes it clear that he considered spectral green to be
uncompounded and not a physical mixture of yellow and blue rays (as
Brewster and Goethe had supposed), but if the mixing of paints is
subtractive, and a mixture of yellow and blue paints, like other green
materials, reflects only uncompounded green wavelengths, how then
does green "equal" yellow plus blue? The only reconciliation I can think
of is that Itten harboured the 19th century idea that colour vision
involves cells "tuned to receive" red, yellow and blue wavelengths (as
Josef Albers would offer as a part of a "plausible explanation" for
afterimages in Interaction of Color in 1963), and that green
wavelengths stimulate the yellow and the blue receptors to create a
mixed sensation of green - an unorthodox view that might have
motivated Itten's statement that "how we discriminate these light
waves is not yet well understood". However it is likely that most
adherents of traditional colour theory instead think of the colours
yellow and blue as residing and mixing in the paints themselves,
though some may also hold completely incompatible scientific ideas
about light and colour in quarantine alongside this idea in their minds.

Figure 2.4.
A. Mixing paths of a set of red, yellow, and blue paints (Cadmium Red
Light, Cadmium Yellow Medium and Cobalt Blue Hue. B. Ittten colour wheel
transformed to show chroma of primary paint mixes. C-H. Spectral mixtures
calculated using the program drop2color by Zsolt Kovacs Vajna.

The other problem is how can primary red paints, if these pass on only
the red rays of the spectrum, subtractively generate full-colour
mixtures with paints extending around two-thirds of the colour wheel?
In terms of modern colour theory they can only make high chroma
mixtures with paints as far away as the nearest subtractive primaries,
yellow and magenta, which share their high reflectance of red
wavelengths. Inevitably, paints of Itten's "primary" red hue make very
high-chroma mixtures with his "primary" yellow, but only very low-
chroma mixtures with his "primary" blue (Fig. 2.4 A,B,E,G,H).
"Primary" blue and "primary" yellow are essentially opposite in terms
of light mixing and opponency, but as we've seen the former
nevertheless reflect enough green wavelengths to cause mixing paths
with yellow paints (all of which reflect large amounts of green
wavelengths) to curve outwards through medium-chroma green (Fig.
2.4A, 1.8G).

Figure 2.5.
A. Recommendation of a palette with two of each (historical) primary
from Parkhurst's The Painter in Oil (1900). B. Illustration of a "double primaries
palette" from The Art Of Color Mixing (M. Grumbacher Inc, 1966).

The impossibility of mixing purples above a very low chroma from

a perceptually pure red and blue paints is a very well-known problem
with trying to use a palette of the traditional primaries. A practical
workaround that dates back at least to the end of 19th century is
the double or split primaries palette, consisting of a "warm" and a
"cool" version of each historical primary (Fig. 2.5). The revival of
traditional colour theory in the 1970's was almost inevitably followed
by the revival of this double-primaries strategy, most famously in the
best-selling book Blue and Yellow Don't Make Green by Michael
Wilcox (1987). Wilcox attached to this traditiuonal double primary
strategy a new formulation of traditional colour theory that
incorporates a model of subtractive mixing without the subtractive
primaries. In Wilcox's 1987 formulation the wavelengths said to be
passed on by a paint correspond directly to its colour, so that for
example a lemon yellow paint is said to reflect a large amount of the
yellow part of the spectrum and a small component or "impurity" from
the green part. Colour mixing of two paints is said to require having a
quantity of this secondary colour impurity or "bias" in common.

Thus while in Itten's colour theory we get orange from mixing red and
yellow paints because "yellow + red = orange", in Wilcox's theory the
red and yellow components neutralize each other, and the mixture
reflects just the orange wavelengths that the paints have in common. In
modern colour theory, all yellow paints reflect most of the red,
orange, yellow and green parts of the spectrum (Fig. 2.4C,D), so
that a subtractive mixture with red paint reflects a large component of
red wavelengths and a smaller component of the wavelengths through
to green, which combine additively to make orange light (Fig. 2.4E),
much as they do in a single-pigment orange paint (Fig. 2.4F). While
each explanation "works" in the context of its assumptions, only
modern colour theory can account for the particular effectiveness of
just one 'impure' blue (cyan), one 'impure' red (magenta) and just one
yellow paint in mixing an optimally large gamut of colours (Fig. 2.6A).
The traditional double primaries palette works in practice because it
includes good subtractive primaries as the three "cool" primaries, while
the three "warm primaries" are useful high-chroma pigments that
access higher chroma reds to orange-yellows and blues outside the
subtractive-primary gamut (Fig. 2.6B).
Figure 2.6.
A. Comparison of approximate gamut of cyan-magenta-yellow (CMY)
primaries with historical red-yellow-blue (RYB) primaries. B. Gamut of "cool"
(=CMY) primaries plus "warm" primaries of a double primaries palette.

Traditional colour theory refers to a scale of value or "tone" as well as

the "colour wheel", but these dimensions are generally either treated in
isolation or integrated only superficially. As a three-dimensional
framework Itten dismissed the Ostwald system, which had been widely
used in other classes at the Bauhaus, and ignored the Munsell system,
which by that time was very widely used in art, science and industry,
and reverted to a simplistic spherical model (Fig. 2.7D-F) externally
like that developed by Runge in 1810 (Fig. 2.7A-C). Itten uses the same
term "saturation" for saturation of colour of light and for relative
chroma of object colours. These spherical models express the way
object colours become more restricted in "saturation" (relative chroma)
as they approach black and white, but their vertical dimension is
essentially meaningless, in that colours of different greyscale value (e.g.
primary red, yellow and blue) are forced to lie at the same vertical level
(Fig. 2.7 G-I).
Figure 2.7.
A-C. Spherical colour order system published by Runge (1810). D-F.
Spherical system published by Itten (1961). G-I. Itten's sphere illustrations
converted to greyscale.

The colour teaching of Josef Albers is sometimes lumped with that of

Itten, but was quite different in many ways. Albers included the
Ostwald and Munsell systems (favouring the latter) in his courses,
ignored Itten's colour circle and sphere, and was skeptical of schemes
of "objective" colour harmony. Albers did treat red, yellow and blue as
the painter's primaries, but he also discussed the psychological and
additive primaries, and his teaching can certainly not be said to be
founded on the historical primaries. Unfortunately but perhaps
inevitably, his much less dogmatic and more experience-based
approach ultimately had less influence on art and design curricula than
Itten's formulaic and self-contained system, which in effect insulates
the teacher and the student from the last century and a half of
developments in our understanding of colour.

The revival of traditional colour theory since the 1960s was not based
on a successful challenge to the truth of modern colour theory, but
must be understood in the context of the concurrent reduction and
elimination of other technical elements, such as traditional drawing
skills, anatomy and perspective, in art education over the same period.
Traditional colour theory anachronistically maintains views of the
nature of colour that prevailed before the late 19th century Helmholtz-
Maxwell-Hering revolution, and its relationship (or lack of
relationship) to modern colour theory is in some ways like that of so-
called "Creation Science" to modern biology. The difference is not
between science and art, but between the sciences of different
eras. Perhaps its most unfortunate effect is that students can become
“adapted― to simplistic treatments of colour, making modern
colour theory seem unnecessarily complicated. If 90 percent of books
treated anatomy or perspective in an extremely simplistic way, you
would get a similar shock the first time you saw some real artistic
anatomy or perspective.
Figure 2.8. A, value 5 plane and B-F, the five hue pages from Albert Munsell's Atlas of
(1915), the forerunner of the modern Munsell Book of Color.
the Munsell Color System

Art and design teachers educated in the age of Itten now occupy
positions of authority and influence, and it is understandable that
many of them are not planning to give up their attachment to simplistic
traditional colour theory any time soon. Nevertheless, there are
encouraging signs that some teachers are beginning to question the
ethics of presenting colour theory that is obsolete in itself and entirely
incompatible with their students' digital studies. However the most
powerful force for change will probably come from students, who pay a
lot for their education these days, and in many institutions have
considerable power to penalize outdated teaching through student
feedback surveys. The current situation among art teachers is
especially disappointing when we recall that a century ago it was, not a
scientist, but an artist and art teacher who published the system that
would become a cornerstone of modern colour theory.
Published June 30, 2015.

Next: Part 1: Introduction