Colours of objects and colours of light

11.18 "Colours of Objects and Colours of Light", CSA National Conference, October 13, 2023


  • 00:00 Introduction: colours of objects and colours of light
  • 04:10 Perceptual layering I: the Red Stripe
  • 13:20 Perceptual layering II: the Orange Cube
  • 17:15 Colours of objects
  • 26:25 Colours of light
  • 27:17 Perceptual layering III: the Banana Split
  • 31:35 Perceptual layering IV: discussion
  • 38:46 ”The colour of the object”

[00:00] This video is a new recording of my presentation for the national conference of the Colour Society of Australia in October this year. It’s a pleasure to thank the committee of the Western Australian Division for putting on the conference and for inviting me as keynote speaker.

[0:20] My presentation builds on some ideas about colours of objects and colours of light from two papers I published earlier this year, combined with some reflections on the reception of those ideas by people ranging from beginning students to various kinds of colour experts. The papers are available online, and you can download them from the link shown here. In this presentation I’ll be using the standard definitions of colour terms provided in the CIE International Lighting Vocabulary or ILV, whose definitions are also freely available online as the e-ILV at the link shown here.

[0:56] The way we see the world has a remarkable layered quality that is often overlooked. Our visual system automatically, instantly and seemingly effortlessly presents us with superimposed perceptions of the varying spectral reflectances of objects, and of the varying intensity and spectral composition of the light falling on those objects. We call these perceptions of spectral properties colours

[1:26] Because we’re mortal objects ourselves, the colours we notice most are those of other objects. Colours of objects can be specified using several different sets of attributes, and although one of these sets characterizes what is called the “Natural Colour System”, I don’t think any of these alternative sets can be convincingly said to be uniquely “natural”. However, for the purposes of this presentation, we can focus on the CIE-defined attributes of hue, lightness and chroma, as illustrated on the left here.

In contrast, colours of lights can be described in terms of the CIE-defined attributes of hue, brightness and colourfulness or hue, brightness and saturation. Because we tend to think mainly of colours of objects, these terms for colours of light are much less widely understood, and even colour educators with decades of experience may present the differences between lightness and brightness, or between chroma, colourfulness and saturation, incorrectly, as being minor and unimportant at an introductory level. But as I want to emphasize in this talk, our perception of colours of objects is inextricably linked with our perception of colours of light. In fact, without considering the attributes of brightness and colourfulness, we can’t even understand the standard CIE definitions of lightness and chroma.

[2:58] These are the formal definitions of hue, lightness and chroma from the CIE International Lighting Vocabulary Notice first that in the definitions of lightness and chroma there is a reference to comparison with “the brightness of a similarly illuminated area”. This expression restricts these definitions to colours of areas viewed in relation to other areas (which are called related colours), and more specifically, to colours perceived as belonging to an illuminated area, that is, perceived as belonging to a light-reflecting or light-transmitting object. By these definitions, lightness and chroma do not apply to colours perceived as belonging to light itself, or to a light-emitting object, unless that object is somehow perceived as being illuminated. (An important example of the latter is a computer screen, which emits light but is normally "read" as an illuminated page). In contrast, the attribute of hue applies to all colours, including those of light-reflecting, light-transmitting and light-emitting objects, or of light itself.

[4:10] Brightness is our perception of the amount of light reaching our eyes from an area, whether that light is emitted or reflected from the area. The lightness of an object is defined as its brightness compared to that of a similarly illuminated white object. If we take an ideal white object to diffusely reflect close to 100% of the light falling on it, the lightness of an object is in effect our perception of the proportion of the light falling on the object that the object reflects. I’ll repeat that for emphasis; the lightness of an object is our perception of the proportion of the light falling on the object that the object reflects. The chroma of an object is in effect defined as the colourfulness of its appearance normalized for the amount of light falling on the object.

I wonder if you’ve ever thought about how remarkable it is that we have a perception of lightness. If we needed to place a perfect diffuse reflector beside an object in order to judge its lightness, it wouldn’t be nearly so remarkable, we could simply compare the amount of light directly. But the fact that we routinely perceive objects as exhibiting a particular lightness without needing to do this shows that our visual system automatically, instantly and seemingly effortlessly presents us with a perception of the amount of light falling on objects, without having any direct way of measuring this amount of light. To do this it must essentially arrive at a guess at the combination of illumination and reflectance responsible for the light reaching our eyes from each area that we see. This computation normally occurs without requiring any conscious intellectual activity from us, though it is presumably based on prior experience of the world. Indeed, our visual system does this so automatically and effectively that it seems to us as if our eyes just directly detect the lightness (and colour) of objects, and this is why it usually doesn’t occur to us how extraordinary this perception really is!

[6:21] The lightness and chroma of an object exhibit a high degree of stability at different intensities of the same illumination. For example, a white object such as a piece of paper is still perceived to be white under illumination varying through a very wide range of intensities, despite the fact that the amount of light reaching the eyefrom the paper varies greatly. This stability of lightness and chroma perceptions demonstrates the effectiveness of the ability of our visual system to present us with consistent perceptions of the amount of light falling on objects.

This high degree of stability of perceived hue, lightness and chroma under varying intensities of the same illumination is one aspect of one of the most remarkable and adaptively useful features of our visual system, called colour constancy. Unfortunately there’s no definition of colour constancy in the International Lighting Vocabulary, but as lightness is an attribute of the colours of objects, I take lightness constancy to be a component of colour constancy, rather than something separate, as it’s sometimes treated.

[6:30] Thus, in this illustration, although the red stripe changes in overall appearance along its length, and is depicted using different RGB values, the stripe itself is perceived to not change its colour; that is, it’s perceived to be uniformly coloured object under varying illumination. So, we judge the red stripe as having the same colour, that is, the same hue lightness and chroma, along its length, and we’d expect it to match the same Munsell chip placed alongside it in the light and in the shadow.

[8:08] But although our red stripe exhibits this high constancy of hue, lightness and chroma under different intensities of the same illumination, the colour appearance of the stripe does vary along its length – it appears brighter and more colourful in the upper part than in the lower part. These colour attributes of brightness and colourfulness describe the appearance of the light reaching our eyes from different areas of the object, as opposed to the colour of the object itself.

So we need a total of at least five different attributes to describe the colour appearance of our red stripe: hue, lightness and chroma to describe the colour of the stripe, and hue, brightness and colourfulness to describe the colour of the light reaching our eyes from different areas of the stripe.

[8:58] A colour perceived as belonging to an object is called an “object colour”. Because they exhibit high constancy through a large range of intensities of the same illumination, the lightness and chroma of an object are perceived as belonging to the object and are thus attributes of the object colour.

In contrast, the variations of brightness and colourfulness here are perceived as being imposed by the illumination rather than belonging to the object itself. Thus, although they are attributes of the total colour appearance of the object, they should not be considered attributes of the object colour.

The colour attributes of blackness and chromaticness used in the Natural Colour System or NCS also exhibit high constancy under different intensities of the same illumination and so provide alternatives to lightness and chroma for describing an object colour.

[9:42] So, we can look at our image on the left in two different ways, as an array of object colours belonging to the stripe and background, or as an array of patches of light of different colours reaching the eye. The first way of looking has been called the distal or constancy mode of visual perception, and is our usual way of looking at the world, in which we are concerned with the properties of external objects. The second way of looking has been called the proximal mode of visual perception. This mode is alternatively called the painter’s mode, because representational painters may view their subject like this and in some way translate the colours of the patches of light seen in this way into the colours of their paints.

[10:34] So, within the image area on the left we have two sets of perceptions overlapping, an array of object colours, and an array of colours of light reaching the eye.

But there’s another colour perception we haven’t considered yet. Within the same rectangle we perceive an arrangement of varying illumination or light and shadow. The colours perceived as belonging to this varying illumination are called illumination colours. Illumination colours can also in hue brightness and colourfulness, but in this example the light is perceived to be achromatic or white light, varying only in brightness.

[11:20] This is an animation from a webinar I gave as the International Colour day 2018 event for the Inter-Society Colour Council, so you can see I’ve been going on about these things for a while. Notice that we perceive these object colours and illumination colours superimposed, as if we see the object colours through the illumination colours. So if someone asks: How can we say that the two ends of the stripe exhibit the same object colour when they look different? We can answer: Our visual system presents the world to us as layered perceptions of object colour and illumination. The object colour, with its relatively constant hue, lightness and chroma, is perceived through the variations in illumination.

It’s only because of this capacity of our visual system to parse visual information into perceptions of object colour and illumination that we perceive objects has having relatively stable object colours and not merely constantly shifting brightness and colourfulness

Because we’re mainly interested in colours of objects, you may not have noticed that our perceptions of object colour are always accompanied by perceptions of illumination, and most places we look we see superimposed object colour and illumination colour peerceptions. The only exceptions, where our proximal perceptions of light are not accompanied by distal perceptions of object and illumination colours, are where we perceive what are called film colours or aperture colours, e.g. a uniform blue or grey sky, sunlight perceived through closed eyelids, or light perceived through a small aperture in a dark screen.

[13:20] Now let’s look at this concept of perceptual layering in a depiction of a three-dimensional environment. In this image we perceive a high-chroma orange cube on a floor of black and white tiles, and we perceive each of these three materials as having the same colour belonging to them in the light and in the shadow. Thus, we perceive the cube itself as having the same high chroma colour on all three faces, as if it were painted all over with the same paint, that we might judge to be about 10R 6/14. At the same time, areas A to C respectively send progressively greater amounts of light to the eye, evoking progressively stronger responses at the level of the retina, and appearing progressively higher in brightness and colourfulness. Similarly, although we perceive the lighter-coloured areas of the floor as being white things, that is as having a uniform white object colour, the corresponding image areas send light of a wide range of intensities to the eye, perceived as a wide range of brightnesses.

[14:35] As always, our perception of these object colours is accompanied by a perception of illumination. In this this image we once again perceive achromatic or “white light” illumination, varying only in brightness, but now we can see that this perception of illumination is rich in information about the three-dimensional form of our objects. Although we might not pay much attention to it consciously, our perception of illumination, supplemented by texture, binocular vision and parallax for close objects, is a major source of information for us on the three-dimensional structure of the world. Notice that once again we experience these perceptions of object colour and illumination colour superimposed in the same rectangle, as if the object colour is seen through the illumination.

[15:38] In the CIE e-ILV definitions of object colour and illumination colour, the word “colour” links to the ILV entry for colour “in the perceptual sense” or “perceived colour” (CIE e-ILV 17-22-040). So, object and illumination colours as defined in the CIE ILV are visual perceptions perceived as belonging to objects and lights. Although they are perceptions, we perceive object colours to be located outside us in objects themselves, as in the uniform black, white and orange object colours that we perceive to be located in the tiles and cube depicted in this image, even though these objects are physically non-existent. This last observation can help students to accept that the colours they perceive to be located in actual objects, are similarly not actually located in those objects, but are perceptions that we project onto objects.

[16:45] If these were actual objects, we could expect them to differ in the property of spectral reflectance, which is the proportion of the light of each wavelength that the object reflects, and for the spectral reflectances of the objects to agree in general character, although not necessarily in fine detail, with the spectral reflectance curves shown on the left here.

[17:15] When we can freely examine an object is daylight, the colour we perceive the object as having is a very reliable indication of its overall spectral reflectance, and conversely, anyone who is familiar with the spectral reflectances of objects can predict the perceived colour of an object so examined from its spectral reflectance.

Or to put this another way: when we can freely examine an object in daylight, the colour we perceive as belonging to the object is the way in which we perceive its overall spectral reflectance, where “overall” means at the level of its long-, middle-, and short-wavelength components, as detected by the human visual system.

[17:47] These provisos are important. When our access to an object is restricted and we can’t freely examine it, the same object can be made to appear any lightness, as has been demonstrated in a number of classic experiments. In the Gelb staircase demonstration on the left, a square that appears black or dark grey in daylight can be made to appear white if it is illuminated in isolation by a bright concealed spotlight in a darkened room. But when a more reflective square is added alongside it, the added square is perceived to be white, and the first square appears light grey. The effect is repeated as progressively more reflective cards are added, so that only the most reflective card appears white, each of the previously white-appearing cards appears a different shade of grey. On the right we see another apparatus used by the same investigator, in which the observer viewed panels in the apparatus from directly above, and the perceived lightness of the panels could be varied by adjusting their slant.

[19:03] Our ability to perceive the overall spectral reflectance of an object as its color is most effective in daylight, the illumination that our color vision evolved under, and tends to perceive as achromatic or white light. Under illumination with a strongly biased or spiky spectral distribution, objects can exhibit different color that may be less representative of the object's over all spectral reflectance. Having said that, our visual system does exhibit a degree of color constancy under spectrally biased, that is, colored illumination. This results partly from adaptation, by which we might perceive the scene on the right to be less strongly reddish than it would appear otherwise, and partly from perceptual layering, by which we might attribute the remaining reddish appearance of the scene to the illumination, so that we might perceive through the reddish illumination some gray and white object colors and other object colors with a degree of constancy.

[20:10] Needless to say, our perceptions of spectral reflectance are not based on instrumental measurements but are shaped by the workings of the human visual system. First in importance is its dependence on the relative responses of three receptor types, so that all that matters is the spectral reflectance at the level of the long, middle and short wavelength components. Thus, spectral reflectances that are physically different but similar at this level (called metamers) may be indistinguishable under some illuminations (as shown on in the middle diagram here), and what is important for the hue and chroma of the colour is the overall direction and amount of imbalance of the spectral reflectance at this level. Also important is the luminous efficiency function, which means that wavelength near the middle of the spectrum have more influence on lightness than reflectance at the two ends.

[21:11] I apologize if I seem to anyone to be labouring the point here and in my papers, but I've done so because a surprising number of my colleagues profess to believe that we don't know what colours are, so it seemed worthwhile to me to attempt a careful statement of what colours and their attributes can be said to be perceptions of, under certain conditions. To sum up, when we can freely examine an object in daylight:

The lightness of thw object is the way in which we perceive the object’s total reflectance of light, that is, the overall proportion of the light falling on the object that the object reflects, as detected by the human visual system.

The hue of the object is the way in which we perceive the overall direction of bias of the object’s spectral reflectance at the level of its long, middle and short-wavelength components, as detected by the human visual system, and the chroma of the object is the way in which we perceive the amount of this bias.

[22:16] Now, if we can’t examine an object freely in daylight, the hue, lightness and chroma we perceive may bear no relationship to the object’s spectral reflectance. There are some who argue, therefore, that colour attributes should be considered purely as perceptions, and that correlating them with spectral properties like this is a naïve error or even “scientism”. But in everyday life we often notice that the colour we perceive an object as having varies under coloured or artificial lighting, and so on, and in such cases it is precisely the colour we perceive when we freely examine the object in daylight that we think of as its “true” or seemingly “inherent” colour, or in short, what we mean by “the” colour of the object. As we’ve seen, colour is a perception, and the concept of a colour actually inherent in an object is illusory, but in communicating about colour in many contexts it would be taking intellectual purity to the point of uselessness to try to prohibit speaking of the colour of an object such as a paint and its connection to spectral properties.

[23:30] I want to briefly refer back to something I talked about at the 2018 CSA conference in Melbourne, because sometimes people have the idea that it's been shown that our visual system doesn't really detect spectral properties, based on demonstrations like this one. Believe it or not, the two trapezoidal gray areas here match physically, as you can see if you block out all the rest of the image. The creators of the demonstration, Purves and Lotto, say in their book that this shows that “what we see deviates from physical measurements of objects and conditions in the real world”, and that's fine as far as it goes – it's acknowledged in a note to the CIE definition of “colour” that colour perceptions depend on the observer and all sorts of things besides spectral properties. But Purves and Lotto claim that their demonstrations are evidence that we need a radical new theory of visual perception. I think it's sometimes debatable exactly what Purves and Lotto are claiming, but some things they say create the impression that our perceptions have nothing to do with spectral properties, and this idea has gained some traction with the general public.

[24:42] But as we just saw, when we can freely examine an object in daylight, its colour is normally a very good guide to its overall spectral reflectance. What Purves and Lotto's demonstrations actually show is that in images of illuminated objects, what we see can deviate dramatically from physical measurements of the image surface. So why is our perception of the spectral properties of this one class of objects so unreliable? Significantly, we don't actually have to block off the remainder of the image to see that the areas match; all we need to do is to introduce a target perceived to be under a different illumination, and we can immediately see that both areas match the same thing. Note that there’s no suggestion that the brightness of either image area changes when the reference chip is introduced; we simply find the relative brightness of the two areas very difficult to attend to until the chip is in place. What I argued in my presentation, which you can find on YouTube, is that we find the perceived colours of the virtual objects depicted in these images so visually insistent that it's very difficult to attend to and compare colour perceptions relating to the image areas themselves. In the context of this presentation, the demonstration shows that as soon as we perceive an area as superimposed object and illumination colours, it becomes extremely difficult to compare it proximally as a patch of light, until we break the representational spell of the image by introducing a target perceived to be under a different illumination.

[26:25] So that's colors of objects; what about colours of light? Rather than spectral reflectance, the colour of a light relates to its spectral power distribution, which is the distribution of power, that is, energy per time, among the wavelengths making up the light.

[26:44] The way I've put it is that the color we perceive as belonging to an isolated light is the way in which we perceive the overall balance of its spectral power distribution relative to that of daylight, as detected by the human visual system. I’ve shown in the first JAIC paper how you can trace this idea back to Newton and forward to the concept of chromaticity in modern CIE colorimetry. It's found in Newton's concept of the "center of gravity" of the "rays" making up a light, which he illustrated with his famous hue circle. This shows this idea that the color of a light containing a mixture of wavelengths can be predicted from the overall imbalance or bias of the center of gravity of the wavelength composition compared to that of achromatic or white light such as daylight. Of the three different lights on the right, the top one is an illuminant representative of daylight, meaning a combination of direct sunlight and skylight, the second is a fluorescent light matching daylight and the third is the emission from a white iPad screen also tuned to match daylight. These three very different spectral distributions are said to have the same chromaticity, meaning they all have the same overall balance at the level of the long, middle, and short wavelength components as detected by the human visual system, and so the three lights all plot at the same point in a chromaticity diagram.

[27:17] So that's colors of objects; what about colours of light? The way I've put it is that the color we perceive as belonging to an isolated light is the way in which we perceive the overall balance of its spectral power distribution relative to that of daylight, as detect detected by the human visual system. So the spectral power distribution is the distribution of power, that is, energy per time, among the wavelengths making up the light.

[28:21] Interestingly it's not it's not a perfectly even spectral distribution that we perceive as being as being white or lacking hue; such an illuminant actually looks pinkish. It's this somewhat uneven spectral distribution of daylight, with a peak in blue green, that we perceive as being white.

[28:44] This is no accident, because it’s to our advantage to perceive typical daylight as colourless, so that we can distinguish objects by their spectral reflectances most easily. What actually happens is that throughout our lives our visual system continually attunes itself so that the somewhat uneven spectral power distribution of daylight is perceived as lacking hue; this adjustment occurs very slowly as our lenses turn brown with age, and over a period of months when we get new ones.

[29:17] I mentioned color constancy before, and as a lot of you probably know, when someone starts talking about colour constancy it won’t be long before they show you a banana!

[29:30] This next diagram is based on a photo that Paul Green-Armytage submitted to the Colour Literacy Project, and I wanted to use it to illustrate perceptual layering using actual objects and lighting. One way it's a bit more complicated than the previous diagrams is because there's another layer involved here: on the plate and a little bit on the banana there's a layer of specular reflection or highlights, classified as luminous color, that forms a third perceptual layer. But setting that aside for now, we see on the left the object colors comprising the light yellow with some darker variants of the banana, the white of the plate and the light brown of the wooden tabletop and on the right you see my representation of the illumination colours. Now today it's possible to generate such a representation using a computer algorithm, but this representation is just my attempt to paint in Photoshop what a white object would look like at any point in the photograph. So you can see you can see it's not entirely achromatic, there's a little bit of reflected yellow light coming off the banana. Once again, we perceive these patterns of object colors and Illumination colors superimposed in the same rectangle.

[30:55] Now here’s something I find very interesting; if we take the RGB color depicting the shaded half of the banana and look at it out of context, we perceive a mid-value olive color under a certain illumination, but in context in the photo we perceive a dimly illuminated light yellow, and it's hard to see that these areas match.  What I think this shows is that, like the Purves and Lotto demonstration, once we perceive an area as superimposed object and illumination colours, it becomes difficult to perceive and compare that area proximally, as a patch of light.

[31:35] Here's a similar example from the early days of my website; area A and area B are the same RGB color, but what you perceive are a dimly illuminated bright red at A and a brightly illuminated darker red at B, that is, combinations of object colour and illumination. Only if I do something like this (introduce a target perceived as being outside the depicted illumination) can you see that the light reaching your eyes from the two areas matches the same thing.

[32:06] And here’s another good one. You perceive the left half of the stripe as a dimly illuminated pink, again as a combination of an object colour and an illumination, and it's very hard to see that the light from this area matches that from the circular area, yet it does.

[32:27] This demonstrates another kind of perceptual layering in which we perceive object colours with a degree of constancy through a translucent layer. On the left we see a Munsell ten-hue circle under a certain illumination, and in the middle  I've emulated in Photoshop the effect of adding a translucent bluish layer, and we still have a perception of the original object colours seen through the translucent layer. But if we remove those rectangular image areas from that context without changing them at all we perceive very different colours. Looking at the middle circle our perception is of red-through-blue, orange-through-blue, yellow-through-blue colours etc, and it’s very difficult to experience colours that relate to the image surface, like we see on the right. But if we take those rectangular image areas out of that context without changing them we perceive very different colours. Looking at the middle circle our immediate perception is of red-through-blue, orange-through-blue, yellow-through-blue etc, and it’s very difficult to experience colours relating to the flat image areas.

[33:22] Perhaps this is debatable, but I think we can discern hints of some of the ideas I’ve been discussing in this passage from the medieval Arabic writer Alhazen, including the idea that an object has its own unchanging colour despite variations in the light falling on the object and the associated variations in the “radiation of colour” from the object, and that “the form of the light” (illumination colour) and the “colour existing in the coloured object” (object colour) are perceived mingled together. The remainder of the passage adds the idea that the separation of these perceptions is learned from experience of the world.

[34:05] The concepts of "discounting the illuminant" and unconscious inference are associated with Hermann von Helmholtz, who discussed them briefly in these passages from his “Treatise on Physiological Optics”. I’ve sometimes seen the idea of discounting the illuminant interpreted to mean simply adaptation; we adapt to the colour of the illuminant and thus see the color of the object. But in the passage from his Physiological Optics he talks about it just after talking about some demonstrations like the one on the left here that involve seeing one color through another color. And what he says is that in the author's opinion the faculty of making such separation depends on this circumstance that we need to be able to do this to separate colors of Illumination from colors of objects. So again, while it's possible to interpret it other ways, I think we could say that he's talking about perceptual layering.

[35:08] Separation of object and illumination colour perceptions is treated explicitly among the Gestalt psychologists like Ademhar Gelb, the Russian born but German trained researcherI mentioned before in connection with the Gelb staircase. In this passage Gelb is clearly expressing the idea that perception of Illumination and perception of object colors are two results of the same process. Skipping ahead to 1978, Barrow and Tenenbaum coined the term "intrinsic images" in the context of their interest in computer vision. They talked about superimposed perceptions of not only reflectance and incident illumination but other things as well like depth, specular reflections and so forth. This research has gone on in leaps and bounds since then, and these are just a few examples from a library of images used to test different computer vision algorithms for how successfully they separate the image into object colors and illumination.

[36:15] Around the time of Barrow and Tenenbaum’s paper, other researchers started talking about how humans might do the same computations. Alan Gilchrist suggested that our visual system distinguishes illumination boundaries and reflectance boundaries, using criteria such as that illumination boundaries tend to be gradational and reflectance boundaries tend to be sharper, and this allows us to build up superimposed images of reflectance and illumination. This research has also gone ahead in leaps and bounds, and I’ll just recommend to you one very helpful paper by Frederick Kingdom that includes an overview of the concept of perceptual layering and the criteria available to our visual system to distinguish reflectance and illumination boundaries.

[37:06] I’ve discussed in the second JAIC paper how colour can play a role in this. In the image on the left I think we tend to perceive uniformly colored stripes under varying illumination, and that’s partly because of the edge properties, which are sharp between what we perceive to be reflectance boundaries and soft at what we perceive to be changes in illumination. But colour is also helping us here; each stripe has a uniform chromaticity, and chromaticity you’ll recall is the overall balance of long, middle and short wavelengths from the point of view of the human visual system.

 [37:43] Now in the processing of information from the retina one line of processing, cone opponency, compares the responses of different cone classes and gives us a perception of the overall balance of the spectral composition. This would allow us to identify patches of the visual field that have the same chromaticity and are therefore likely to be made of the same stuff, and therefore any change in the perceived intensity of light within these areas is likely to be due to a change in illumination. So, if you get a number of those changes of Illumination arranged consistently with each other it helps the visual system to guess that a change in illumination that's occurring, but if the arrangement is not consistent then it doesn't work.  These are exactly the same rectangles, but they're just turned around so that the gradients are not consistently arranged, and so they don't create a perception of consistent illumination.

[38:46] I'd like to finish by sharing something I've noticed about the reception of these ideas with different audiences. I’m privileged to be able to discuss these ideas with people ranging from people just beginning think about colour to various kinds of colour experts, even within my online colour course, as well as in online and in-person discussions, and I've also sought feedback from the other experienced colour educators of the Colour Literacy Project, and from some world authorities on the science of colour appearance. And it seems to me that people fall into three fairly distinct groups, correlated with what they understand by "the colour of an object".
For the colour novice the concept of an object colour as defined by the CIE, presents no problem; bananas are light yellow, with maybe a bit of green and black. This light yellow colour can be described as a certain hue, lightness and chroma, and is closely related to the banana’s spectral reflectance. And equally the concept of colour constancy is unproblematic; bananas are light yellow whether we look at them indoors or outdoors in sunlight.

[40:01] Beginner painters tend to paint objects in this object colour, ignoring or reducing variations in appearance due to light and shade. This is very noticeable in many children's paintings, and it's also characteristic of much ancient painting including Egyptian, Minoan, classical Greek, and Etruscan painting.

[40:29] But as we learn to paint we have a lightbulb moment where we learn to attend to the complete colour appearance of our subject, including the influence of light and shade, specular reflections, and reflected light from other objects, as in this lovely demonstration by James Gurney. The same lightbulb moment occurred in the evolution of ancient Greek painting between the classical and Hellenistic eras, as documented in Pliny the Elder’s Encyclopedia. It was largely lost sight of in the Middle Ages, and then recurred in the European Renaissance.

[41:05] Experts in every colour discipline go through a similar lightbulb moment at an early stage of their education, and become aware of the wide range of factors contributing to colour appearance. As a result, many come to think of “the colour of the object” as meaning not simply the object colour as described here, but the total colour appearance of the object, including the colour variations of the light reflected from different areas of the object. By this view, brightness and colourfulness are included as attributes of the “colour of the object”. This makes the relationship of this “colour of the object” to spectral properties complex and indirect, which may contribute to the confusion I’ve noted about what colours are. The“colour of the object” in this expanded sense does not exhibit constancy at different intensities of the same illumination, and so that the whole concept of colour constancy may be rejected, or restricted to the weaker constancy exhibited by objects under spectrally different illumination.

[42:12] The “colour of the object” in this expanded sense can be quantified by the values of the pixels in a photograph of the object, and there have been some attempts recently to investigate object colour perception by statistical analysis of CIE Lab values from photographs. But we should remember that a photograph is a physical artefact created by a photographer using settings chosen according to the purpose of the photograph. If the primary aim of the photographer is to accurately depict the colours of an object, she will choose a white balance that makes the illumination appear neutral, which results in the pixels depicting the object exhibiting much the same hue as the object, but usually a substantial range of lightness and chroma to depict the object under different levels of the illumination. In addition, with the same aim the photographer will tend to choose an exposure such that the lightest pixels in the hue series match the lightness of the object, or are perhaps just slightly darker to allow some “headroom” for specular highlights. It seems to me that the results reported from these studies tell us the properties of a well-exposed photograph rather than properties of colour vision. What we need to know is how the light from an object is anchored to a particular object colour/ reflectance, when we don’t have the exposure chosen by the photographer doing that for us.

[43:44] Returning to our different audiences, it can take another lightbulb moment to realize that this second viewpoint has lost sight of the idea, usually evident to the colour novice, that we perceive objects as having a relatively stable colour belonging to them. This third viewpoint recognizes that the total colour appearance of an object is layered, and consists of superimposed perceptions of (at least) object colour and illumination colour. This object colour is the way in which we perceive the spectral properties of the object, its spectral reflectance, and this illumination colour is the way in which we perceive the spectral properties of the illumination, its intensity and spectral composition. And it’s this object color that we perceive through the illumination that exhibits a high degree of colour constancy under different intensities of the same illumination, along with generally weaker constancy under spectrally different illumination.

[44:45] Sometimes if a student tells me in class that they’re confused, I begin my answer by saying “don’t worry, you were probably confused before, you just didn’t know it!”. I hope that by understanding that we can mean quite different things by such deceptively simple expressions as "the colour of the object" and "colour constancy“, it may help us to negotiate some of the unrecognized confusions that plague our discussions about colour.

I said at the start that as mortal objects, the colours we mainly notice are those of other objects, and because of this the attributes of object colours are far more familiar than the attributes of colours of light, and most colour educators pass on the idea that there are just three attributes of colour. The solution is to begin colour education with perception, through “eye-opening” exercises like those of the Colour Literacy Project that address how our perception of the world involves colours of light - both light reaching our eyes and light falling on objects - as well as colours of objects.

Links mentioned in presentation:

Briggs, 2023b. The Elements of Colour. Parts One and Two. Journal of the International Colour Association, Volume 33, Special Issue: Contributions of the Colour Literacy Team, editor Robert Hirschler:

Commission Internationale de L’Eclairage (2020), CIE S 017:2020 ILV: International Lighting Vocabulary (2nd edition). Definitions available online as the e-ILV at

Attributes of colours perceived as belonging to (A) objects and (B) light.

Alan Gilchrist’s ECVP 2014 Perception Lecture, Theoretical approaches to lightness and perception.

Colour & Vision Research Laborarory, University College, London

R. Beau Lotto

Briggs, D. J. C., Colour Constancy Illusions and Painting. CSA 2018 conference:

James Gurney, Banana demo, gouache, 2014

This page published February 7, 2024.
< The Elements of Colour