The Dimensions of Colour Today

11.13 The Dimensions of Colour Today

This video is a recording of a presentation I gave on March 20, 2022, as part of the Colour Society of Ausralia's International Colour Day 2022 event, "Progress in Colour Education".
[2:00] A limitation of a lot of colour education in art and design is in its treatment of the attributes or dimensions of colour. “Colour” is often considered in isolation from lightness or “tone”, and presented in terms of a single two-dimensional hue circle or “colour wheel”, very often in a form embodying historical beliefs about three “primary colours”. Even when colour is considered three-dimensionally, it’s usually with the assumption that a single set of three colour attributes suffices for all purposes and for all modes of colour appearance, and that these three attributes can be adequately represented by a simple symmetrical model like a sphere. It's been nearly 15 years since I first addressed these issues in my website The Dimensions of Colour, and in this talk I’ll give an overview of the main alternative frameworks available to painters today for visualizing colour relationships, and highlight some positive developments stemming from a number of sources that have emerged over this period.  
[3:10] For painters using physical media the most widely used three-dimensional framework is the Munsell system, which classifies colours of paints and other physical objects in terms of the colour attributes of hue, lightness (called “value” in the Munsell system) and chroma or strength of colour.  What we’re seeing here is the range or gamut of paint colours that can be obtained from mixtures of a particular set of artist’s acrylic paints. When we look at what we might call the individual hue pages we can see that the colours reach their maximum chroma at different values for different hues, forming an irregular solid that can’t be reduced to a regular sphere without either distorting the chroma dimension or leaving out many paint mixtures. Painters learn to think of paint mixtures as following paths through this space, and it’s essential for them to understand that the space is irregular and varies in extent for different sets of paints.
[4:24] A very useful procedure for using this framework to match a target colour in paint is to select paints either side of the target colour in hue, and to raise or lower them to the value of the target colour, before mixing these to obtain the required hue and then if necessary reducing the chroma by adding a neutral grey paint of the same value.
[4:56] These examples are screenshots obtained using ZKV’s program drop2color which generates 3D plots of paint mixing paths and gamuts in Munsell space. This free program became publicly available not long after the launch of DOC, and it’s a wonderful resource for helping students to visualize mixtures of paints as following paths through hue-value-chroma space. This short movie will give you a feel for the three-dimensional output of the program.
[5:43] However all young art and design students today also need to understand digital colour. Munsell notations can be converted to and from digital colour specifications but only by going through a table rather than directly through a formula. For digital painters and in the digital environment generally it’s more convenient to use the colour space CIE Lab, known as Lab space in Photoshop. CIE L*a*b* was designed to arrange digital colours similarly to the Munsell system, although it specifies them in terms of the reddish vs greenish and yellowish vs bluish chroma components a* and b* instead of hue and overall chroma.
[6:29] Another very useful tool for teaching colour is Color Inspector 3D, which can be used to display colours from a photograph in CIE Lab and other colour spaces. It’s actually been around since 2004 although I only came across it a few years ago. Again, this short movie will give you a feeling for the three-dimensional output of the program and how it can be used to help students to visualize paint mixing paths in a hue-value-chroma space.
[7:07] So for example we can compare the gamut or range of mixtures obtained from different sets of paints, here we see plotted the changes in hue and chroma as well as value when we add white paint to two colourant paints. And here we see the mixing path of two paints of different hues but similar values, showing the typical drop in value that you may have also noticed earlier in many of the mixing paths in drop2color.
[7:55] So, lightness and chroma provide one way to classify colours of objects including paints, and to divide up a hue page of colours like this, where lightness increases going up the page and chroma increases to the right. But there’s another way we can think of variations of colour on this page: We can consider these colours to be perceptually made up of black, white and pure colour components. So towards the top corner the colours are perceived to have a high white component, towards the bottom corner they’re perceived to have a high black component, and towards the right corner they’re perceived to have a high coloured component.
[8:37] This is the basis of the Scandinavian Natural Colour System or NCS, in which these components are assigned numbers taken to add up to 100. Colours are specified according to their perceived black content (called “blackness”) and their perceived colour content (called “chromaticness”, or in standard CIE terms, chroma relative to the maximum chroma possible for the hue).
[9:08] If we were to arrange the triangles for every hue around a central axis we’d get a symmetrical colour space in the form of a double cone, but as you can see the actual coloured chips do not fill the triangles and occupy varying parts of the triangle for different hues, so that like the Munsell system the range of actual paint chips forms an irregular solid.
[9:36] Something that can be of great interest to us as painters is what happens when we cross that upper line of zero blackness. If we were to take any chip and increase the amount of light that it reflects or emits without changing its spectral composition, the colours we perceived would move along a line radiating from the black point, like the blue line here. Blackness would decrease to zero at the upper line, and then the chip would begin to appear fluorent, that is, to show the appearance of physical fluorescence, the appearance of being unnaturally bright, before reaching the appearance of luminosity, the appearance of a light-emitting object. This sequence of states of appearance, passing from black through decreasing blackness to zero blackness, followed by fluorence and then luminosity, is called brilliance.
[10:36] In this demonstration from my paper at AIC 2021 in Milan I’m going to progressively increase the amount of light the circular area on the right side of the cube emits relative to its surrounds. With little or no light emitted the area is perceived to be black; at this level the area appears brown, a colour exhibiting a certain degree of blackness; at this level the area appears a bright orange of approximately zero blackness, that is it exhibits neither blackness nor fluorence. Now if I increase the light further I get this unnaturally bright or fluorescent appearance called fluorence. Interestingly the circle on the top face is depicted with exactly the same RGB colour, but does not appear fluorent in this context seen as more brightly lit it. I can’t turn up that orange light any further, so to show the next step I need to evoke a more dimly lit context [click], and now the same RGB colour gives the appearance of a primary light source. Understanding these relationships can obviously be very helpful for painters if they are concerned with creating the appearance of luminosity and consistent illumination.
[12:12] The NCS has published a key of NCS notations for Munsell chips when viewed on a light background, and it’s possible to use these tables to draw in contours of equal NCS blackness on Munsell hue pages as I’ve done here. These pages show that the point of zero blackness lies close above the top of the Munsell solid, near or just above the lightest chips of each hue and chroma. We can see that the point of zero blackness occurs at different Munsell values, depending on the hue and chroma. Zero blackness and the transition to fluorence occurs at high values for all pale colours and lower values for strong colours (at the lowest values for strong blues and violets, moderate values for strong reds and greens, and light values for strong yellows).
[13:13] This concept is vital to fully understanding colour relationships on hue pages like these. The Munsell system is designed so that chips of the same Munsell value reflect light of the same luminance under a standard daylight illuminant, where luminance is a measure of energy of light from the point of view of the human visual system that factors in the very different responsiveness of this system to light of different wavelengths.
[13:41] This corresponding diagram in CIE L*a*b* shows the gamut of digital colours rather than physical paint chips, and has a middle grey rather than a light background. These two factors mean that the highest-chroma swatches begin to encroach into fluorent territory- they have a bit of a glowing look about them. As in the Munsell system, all swatches in the same horizontal row emit light of the same luminance, If we look at a horizontal row of chips, such as the value 5 or middle grey value row, then as we pass outwards along this row from the grey axis the Munsell value and thus the luminance stay the same, but the brilliance or freedom from blackness increases. This increasing brilliance can result in an overall impression of brightness that can cause people to estimate the lightness of high-chroma chips to be greater than their Munsell value or their lightness in CIE Lab, and the glowing look of the highest chroma swatches might lead untrained observers to estimate them to be higher in lightness than 50, perhaps 60 or 70. There’s a view that this is a fault in Munsell value and CIE Lightness as measures of perceived lightness, but I would say that two distinct perceptions are being confused here. There is a perception of luminance-related lightness that shows up as distinctness of boundary: note that in the value 60 row all of the swatches have a similarly distinct boundary with the value 50 grey background, and that in the value 50 row all of the chips display a very indistinct brightness contrast with the background, despite the fact that some of the chips have a glowing appearance.
It may help to place the value 60 grey swatch adjacent to the glowing red swatches. I hope you’ll agree that when we examine the border between the grey swatch and the red swatches, the grey swatch is the lighter of the two, even though the latter appears glowing. So I think there are really two perceptions here, luminance-related lightness and this “glowing” impression associated with high brilliance.
[15:37] So I hope I’ve convinced you that to know your way around a hue page it’s useful to understand how it can be divided according to blackness/brilliance as well as according to lightness. In teaching painters I would start with the lightness-based systems Munsell and CIELab because of the enormous importance of lightness for painters of all kinds, but the important thing is that both approaches should be understood before too long. A great resource is the Munsell hue page sorting exercise included in the NMSCS, and Steve Westland of the Colour Literacy Project has created interactive online versions of this exercise that allow you to drag and sort virtual colour chips on a hue page. The NCS has produced chip-sorting exercises illustrating the concepts of white, black and chromatic content, and Paul-Green-Armytage and Maggie Maggio have devised chip sets designed on a similar basis intended for young children for the Colour Literacy Project.
[16:45] So lightness and blackness provide two different ways of dividing up a hue page of colours of paints and other objects, but neither are applicable to colours of light, such as light perceived as coming from a primary light source. I’m hoping you’ll perceive the black panel in middle of this illustration as having an array of nine luminous dots. If so, you’ll see that when we’re thinking of these dots as lights we can’t describe them in terms of lightness or blackness, they’re off the scale of lightness and off the scale of blackness.
[17:22] We need another set of terms to describe colours of lights. One way we could describe them would be to say that the three on the left are pure white lights, varying only in brightness, the three on the right are all pure orange lights, varying only in brightness, and the three in the middle all exhibit a similar balance of white and orange components, varying only in brightness. This perceived proportion of a coloured light component, or freedom from a white light component, of a light is called saturation.
[18:01] Now as the brightest white light is brighter than the brightest whitish orange light, and the brightest whitish orange light is brighter that the brightest saturated orange light, we could classify the nine lights in terms of brightness and saturation like this.
However, you may notice that the middle and right-hand series of lights become more colourful as they become brighter, their strength of colour increases, so another way we could classify these lights is according to brightness and colourfulness like this. The saturation or purity of colour of the lights is now shown as the angle between the series and the vertical axis.
[18:47] These terms brightness, saturation and colourfulness also apply to colours of light reflected to our eyes from different areas of a scene, and so can be used to describe the appearance of a scene: for example, we might say that the lighter coloured areas of the floor appear to have the same high lightness (to be white things) but vary greatly in brightness (amount of light they are perceived to send to the eye).
[19:20] The lines of uniform saturation turn out to be important in that the light reflected to the eye by an opaque uniformly coloured object under different levels of illumination would be expected to follow these lines. For example, our three saturated orange lights correspond to the light reflected to our eyes from the left, right and top planes of this cube, and our three less saturated orange lights correspond to the light reflected to our eyes from the same planes of this cube. To paint the appearance of this scene we would in some way translate these differences in brightness and colourfulness into differences in the lightness and chroma of our paints.
[20:13] The attributes of brightness and saturation help us to properly understand a very widely used digital colour space, HSB, which considers pixel colours as colours as lights. S is a simple index intended to predict the saturation of the digital colour relative to the maximum saturation possible for that hue, and B is intended to predict the brightness of the colour relative to the maximum brightness possible for digital colours of that hue and saturation.
[20:48] On its right hand side the Adobe Photoshop colour picker also provides an alternative way of specifying digital colours, as object colours having lightness, using Lab space.
[21:03] Blackness-based specifications are not as readily available in the digital realm at present, but a complete set of NCS colours can be downloaded as a free swatch file from the NCS website, and a colour space specifying digital colours in terms of relative whiteness and relative blackness has recently become available in CSS.
[21:27] Just as different three-dimensional colour frameworks are useful for different purposes, different hue circles may also be needed for different purposes. A minority of painters refer a physical colour atlas (usually the Munsell Book of Color) to specify paint hues, but many more make use of one or more hue circles as a conceptual framework. I’ll briefly outline the main hue circles that painters today think in terms of:
[21:58] The Munsell hue circle is structured around five opposite pairs of hues, so it’s a bit more complicated to learn than the three opposite pairs of many popular colour wheels, it but it has the advantages of being perceptually even and of showing to a good approximation additive or visual complementaries, for example showing red opposite turquoise (blue-green) and green opposite magenta (red-purple).
[22:58] When a simpler conceptual framework is all that’s required, a basic hue circle structured around the unique hues arranged in the opponent pairs yellow opposite blue and red opposite green may suffice. Neither the precise hues deemed to be “elementary” red, yellow, green and blue in the NCS, nor the slightly different hues conventionally regarded as “middle” red, yellow, green and blue by painters, are evenly spaced perceptually, but this may not be a significant drawback for many purposes.
[23:03] The appeal of the traditional red, yellow and blue primary colours is understandable in the context of the concept of unique hues, which holds that red, yellow and blue, along with green, are perceived to be the pure components that combine to form intermediate hue perceptions. As I argued last year in my presentation for Colour Connections, the omission of green from the traditional primary colours can be seen as a relic of the view that prevailed before we understood subtractive mixing, when the colour green was thought to be physically composed of yellow and blue. Omitting green and placing middle red, yellow and blue at equal distances results in a hue circle that is perceptually very uneven and in which some of the opposing hues are far from visual complementaries.
Despite these substantial drawbacks, some very skilled painters continue to favour the traditional colour wheel as a simple practical framework, although one suspects that an equally simple but less problematic framework might serve their purposes even better. The most unfortunate aspect of the continued use of the RYB framework is that it encourages misconceptions about colour itself. It can take considerable effort to dislodge from a student’s mind the idea that a greenish yellow “contains some blue” or a greenish blue “contains some yellow” once it’s in there.
[24:34] If we replace the middle red and middle blue of the traditional colour wheel with magenta and cyan respectively, we can obtain a hue circle that is less distorted in perceptual spacing and in complementary relationships. Further to this, a cyan, magenta and yellow palette has a larger and more even gamut than other sets of three colourant paints, including the three traditional primaries red, yellow and blue, but what’s more important is the reason why this is so, a distinct difference in the pattern of mixing paths for paints close to cyan, magenta and yellow hues compared to paints distant from these hues.
[29:34] In most graphics programs hue is measured by hue angle (H) in HSB colour space, which is very perceptually uneven, in that the same angular change corresponds to a much smaller hue change in some sectors (especially the green and blue sectors) than others. Opposite hues are complementary on the 0-180, 60-240 and 120-300 degree axes, but not on intermediate axes, as explained in Dimensions of Colour.
[30:08] Although CIE L*a*b* was designed to specify colours in terms of a* and b* components, these can readily be converted to predictors of chroma and hue angle in the format L*C*h. This measure of hue angle is much better than hue angle in HSB, but so far is not widely available in graphics programs apart from the open-source program GIMP.
Many painters using traditional paint media, especially watercolours, have become familiar with specification of paint colours in terms of CIE L*a*b* and the colour appearance model CIECAM02 through the very impressive and highly influential Handprint website of Bruce MacEvoy, created in 1998 and continually extended since then.
[30:59] I want to end by emphasizing the positive developments I’ve seen since I posted the first pages of Dimensions of Colour in 2007. Very soon afterwards Graydon Parrish with others created the Rational Painting forum, which not only brought together numerous skilled exponents of the Munsell system as applied to painting, but was also a home for high-level discussion and exchange of information about all aspects of our current understanding of colour. Two other contributors to RP who many of you will know from our conference last year were Ron Francis, who paints vividly atmospheric compositions from his imagination by applying the science of light and colour, and Andrew Werth, who uses his deep understanding of colour perception to create striking abstract paintings.
[31:29] Graydon Parrish, Richard Murdock and other former contributors to Rational Painting, notably Douglas Flynt and Paul Foxton, have gone on to contribute very positively to the understanding of colour among painters through their live and online painting classes, their blogs and Facebook groups, and/or their online videos.
[31:53] There’s also James Gurney’s excellent book Color and Light, from 2010, which deservedly has a very strong reputation among young students of painting. The book provides an excellent introduction to a broad range of topics including different hue circles, physical aspects of lighting, visual perception for artists and much more.
[32:18] I can’t think of a more positive note to end on than the work of Peter Donahue, who posts on TikTok as color.nerd. In a short time he’s attracted a huge following with his entertaining and well-researched posts on modern colour theory for painters. The thing I love about Peter is that I’m sure people tell him that things like Macadam ellipses are too advanced for a general audience, but he just goes ahead and makes a video about them and gets 25,000 views.
I think it’s fair to say that it’s become harder today for people to teach traditional red-yellow-blue colour theory without someone in the class or in the comments section challenging the teacher based on these newer sources of information. I remain unsure to what extent modern colour theory will replace traditional colour theory in formal education. At the moment it seems more that the old colour courses are simply dying out, and students are increasingly getting their information from other sources. Of course not all sources are as well-researched as Donahue and Gurney, far from it, so one thing we can do in the Colour Literacy Project is to help students to find the best sources.
Resources mentioned in the introduction:
Understanding and Applying Colour
This public programs course of eight weekly 3-hour classes is offered live on Zoom through the National Art School, Sydney, four times each year beginning in February, April/May, July and October.
Colour Society of Australia
Colour Literacy Project
AIC Study Group on Art and Design
Resources mentioned in the presentation:
Colour Mixing Tools including drop2color (Zsolt Kovacs-Vajna)
Color Inspector 3D (Color Inspector 3D plugin for ImageJ)
Color Atlas (Harald Brendel)
More than Three Dimensions: Communicating the Attributes of Colour Perception in Colour Education (my AIC Milan 2021 presentation)
Four Key Insights about Colour (my Colour Connections presentation)
NCS swatches for Adobe CS/CC
Hue, whiteness, blackness in CSS
Bruce MacEvoy
Graydon Parrish
Richard Murdock
Paul Foxton
Douglas Flynt
James Gurney
Peter Donahue

Other excellent resources:
Artists' Helper (Bob Burrage)
Virtual Atlas (VCS Consulting)
Virtual Munsell Color Wheel (Andrew Werth)
ColorWell (John Morfis)

This page published March 31, 2022.
< A Shillito Student Poertolio from the mid-1940's