Colours as Observational Properties

The second main claim made by the naïve realist is that colours are distinct from the physical properties of objects.

In saying that colours are distinct from the physical properties of objects, the naïve realist is not necessarily saying that are ‘perfectly simple’ properties whose nature cannot be described further; indeed, on the face of it this is inconsistent with the claim, outlined in yesterday’s post, that colours are mind-independent properties. (It is partly for this reason I prefer to call the position ‘naïve realism’ rather than ‘primitivism’.)

Rather, to say that colours are distinct from the physical properties of objects is to say that they cannot be reductively identified with properties that can only be described using vocabulary from the physical sciences. So, for instance, colours are not dispositions to reflect light in different proportions across the electromagnetic spectrum (‘surface reflectance profiles’), or microphysical properties of objects, as they are according to common contemporary forms of colour physicalism.

I spend three chapters of A Naïve Realist Theory of Colour defending the claim that colours are distinct from the physical properties of objects. One line of argument for this claim is grounded in the idea that our colour concepts are concepts of observational properties.

A property is an observational property when it is not possible for something to look to normal perceivers in normal conditions to have the property, but not in fact have it. The property of being an apple, for example, is not observational in this sense, because whether something is an apple depends upon properties that cannot be perceived. It is possible for something to look like an apple to normal perceivers in normal conditions, but not be an apple: for instance, if it is a perfect plastic replica of an apple. The same is arguably not true of properties like colour or shape. If something looks yellow, or looks spherical, to normal perceivers in normal conditions, then it tempting to say that it must be yellow or spherical. It isn’t clear that we can make sense of the idea of ‘fool’s colours’ or ‘fool’s shapes’: things that share the appearance of colours or shapes, but which aren’t really colours or shapes because they differ in their non-perceptible properties.

One way of presenting the general challenge to the colour physicalist is as follows: to square the claim that our colour concepts are concepts of observational properties with the claim that colours can be reductively identified with properties with a complex physical essence.

Consider, by way of illustration, a form of colour physicalism that identifies colours with types of ‘surface reflectance profile’. Most physical objects reflect light in different proportions right across the electromagnetic spectrum, and what determines the colour they will appear is the proportion of light that they reflect at different parts of the electromagnetic spectrum that is visible to humans, between roughly 400 and 700 nanometres. This is called their ‘surface reflectance profile’.

Colours cannot be identified with individual surface reflectance profiles, unless there are many more colours than we normally assume there to be. Objects that differ in their surface reflectance profile—often quite radically—can nevertheless appear identical in colour (at least under certain conditions). This is a phenomenon known as ‘metamerism’. Our reaction—or more specifically, lack of reaction—to the discovery that objects that differ physically can be identical in colour reflects, I suggest, the fact that our colour concepts are concepts of observational properties. Even after we have discovered that objects that appear identical in colour can differ physically, we nevertheless continue to treat them as identical. In this respect, our judgments about colours do not display a deference to science in the way that judgments about whether, say, whales are fish do.

Contemporary physicalists like Alex Byrne and David Hilbert accommodate the phenomenon of metamerism by identifying colours, not with individual surface reflectance profiles, but with types of surface reflectance profile that all appear identical in colour. But it seems possible to imagine that there could be objects that differed still more radically in their physical properties – for instance, perhaps they acted directly on the visual cortex – but which still appeared to be coloured. My sense is that we would treat these objects as coloured, too. The reason for this, I suggest, is that we ordinarily take colours to be observational properties that form a largely autonomous domain. Whether or not a colour ascription is true depends on the way an object appears to normal perceivers in normal conditions, not on its unobservable physical properties.

In the book, I develop this argument via a variation on Kripke’s modal argument against physicalist theories of mental states like pain. I also provide responses to two lines of objection to the claim that colours are distinct from the physical properties of objects: that colours understood in this way can play no role in causing colour experiences, and that there are no properties of physical objects that instantiate the appropriate structural properties of the colours (for instance, properties which stand in the right relations of similarity). These are interesting and important issues. In my final two posts, however, I want to step back from debates about the nature of colour and try to give a sense of some of the more general issues that thinking about colour can help to illuminate: questions about the nature and possibility of philosophical inquiry and the problem of consciousness.



  1. Thanks for an interesting post Keith.
    Shapes are observational properties, on your account, and this means that it is not possible for something to look to have a certain shape, to normal observers in normal conditions, yet not in fact have it. I assume that the same holds for lines (because shapes are made of lines). What would you say about the Muller-Lyer illusion? A line appears longer than another to normal subjects, in normal viewing conditions (lighting is normal, etc.), but it is not in fact longer. Consider also Shepard’s tables illusion, in which two identical parallelograms appear to have different shapes (one long and thin an the other diamond-shaped). Would you say that the conditions in these cases are abnormal? Doesn’t this run the risk of trivializing the idea of normal conditions, so that, by definition, any condition leading to a non-veridical experience of shape counts as abnormal?

    • Keith Allen

      It’s a good point. You need to understand ‘normal conditions’ as involving a suitable range of normal conditions, since visual illusions like the Muller-Lyer often work by exploiting the fact that a stimulus seen in one set of conditions will look the same as a stimulus seen in another. Hopefully this doesn’t trivialise the distinction with non-observable properties, however, as even across a suitable range of normal conditions it possible for things to appear to have these properties when they don’t.

  2. Thanks for the really interesting summary of you argument. I’m very much enjoying it. I’d like to question one statement though. You say:

    “If something looks yellow, or looks spherical, to normal perceivers in normal conditions, then it tempting to say that it must be yellow or spherical. It isn’t clear that we can make sense of the idea of ‘fool’s colours’ or ‘fool’s shapes’: things that share the appearance of colours or shapes, but which aren’t really colours or shapes because they differ in their non-perceptible properties.”

    We surely need to distinguish here between perception of surface colour (a material property that can be perceptually estimated) and colour experience (a phenomenal property). When we look at two patches of material cut from the same piece of cloth, one in shade and one on sunlight the two colours we experience when looking at them are different – the one in shade elicits a duller, darker, colour experience. We can, nevertheless, judge correctly that they are made of the same coloured material. As you say “Properties that exhibit perceptual constancy, like shape, size, and colour, manifest a distinction between appearance and reality: between the way objects instantiating the property (merely) appear in particular conditions, and they way that they really are – and, when everything goes well, can be perceived to be.” I would suggest that it is quite possible for something to look yellow, in the sense that we perceive it to be made of yellow material, when it is not, in fact yellow. In fact I would even argue that this specific example is a pretty common experience! When you look at a tree illuminated by dappled sunlight it is quite easy to confuse patches of sunlight for patches of yellow leaves. So we perceive yellow leaves where the leaves would normally be perceived because we have mis-estimated the nature of their illumination. There remains an interesting question as to whether the yellow we experience as a colour sensation changes because we have mis-perceived the material colour of the leaves.

    • Keith Allen

      Hi Bob, thanks for the comment. I think you are right that we can distinguish surface colours (that are illumination-invariant) and apparent colours (that vary with the illumination). In the case of the cloth you mention, for instance, it is possible to attend to either. But I think there’s a further question here about exactly what these apparent colours are. One option, as you suggest, is that they are phenomenal properties of experience. The view I suggest in the book is that they are actually relational properties that coloured objects have in particular conditions. One reason for this is that the duller, darker colour-related property of the cloth-in-shade appears to be ‘out there’ in the world just as much as its surface colour. These apparent properties may, in turn, be able to explain the kinds of error that you mention: perhaps what is happening in these cases is that we mistake one type of property (an apparent, variable colour) for another (a constant colour). That said, I think that this particular explanation of differences in experience as the illumination varies is independent of the claim that we perceive constant colours, which is the key claim that I want to defend.

  3. Raul Villuendas

    You’re partially right. Colors are distinct from physical properties but I’m not fully aligned with your statement that affirms colors are “concepts of observational properties”.
    I think the idea would be better explained saying colors are concepts that emerge from our observational “SYSTEM”.
    It is our visual and cognitive system that generates the concept of color. So as any physical complex system of information management it generates concepts that help interacting with objects and their inherent physical properties.

    Like we see today, complex systems like a convolutional neural networks ( generate concepts like the “circolarity”, the “spirallity”, the eye, mouth, and color as well (redness, yellowness, etc.) in case of visual recognition neural networks.
    So concepts related to the visual recognition emerge within our complex visio-cognitive system when interacting with the object. And then they become available for other cognitive system in our brain to use them (like language).

    This explains for example that different visual systems like the one from bees generate a different concept of color as their cognitive system interact differently with the property of objects of reflecting the electromagnetic spectrum.

    This makes clear as well the ontology of qualia, demonstrating that qualitative properties of things do emerge from the auto-organization of complex systems, what is to say, they emerge from auto-organized-quantity.

    Work of Stuart Kauffman and recent books like “Auto-organizzazioni” by Alberto F De tOni, Luca Comello and Lorenzo Ioan can better illustrate the idea.

    • Keith Allen

      Thanks for the comment Raul. I agree with you that the way that we perceive colour depends, in part, on the way that our visual system works. But we disagree, I think, on whether this undermines some form of colour realism. The way that we perceive anything depends, in part, on the way that our visual system works, but this isn’t normally taken to show that, for instance, shapes don’t exist (although some people do accept this). What I’d like to suggest is that we can think of visual processes mechanisms as making the environment available to perceived; in the case of colour this may be a matter of determining which aspects of the environment we perceive, which is why colour perception differs between members of different species. I’ll suggest in my final post on Friday how this provides a different account of the qualitative nature of experience, one that doesn’t appeal to qualia.

  4. Raul Villuendas

    Looking forward to your post on friday.
    Nevertheless, I would add that shapes do not exist for bats but of course they do for us.
    Shapes, colors, etc. are properties that are part of our ontologically subjective world (quoting Searle) and they do exist but they do “emerge” of the constant collision of 2 systems: what we call objects (ontologically objective world) – and what we call perceptive-cognitive systems (these ones emerged as part of the laws of evolution). Memory then make these concepts or properties available for further use in our cognitive system (this was one of the breakthrough innovation of evolution).
    If we would observe the world through a powerful electronic microscope we would not develop the concept of color (like bats or blind people) )and our concept of shape would be much more fluid and fuzzy as particles would not respect the Pauli exclusion principle.
    So, this is to say, properties of things are not absolute or objective entities but completely relative concepts contingent to our senses, to our existence. We can only talk abouot the “state of affairs” through our sense so through our concepts but they remain ontologically subjective to us.

  5. While I would certainly agree that “Colours cannot be identified with individual surface reflectance profiles, unless there are many more colours than we normally assume there to be. ” I would argue that, if we insist on assigning colours as properties of objects, then yes there are many more colours than we normally assume there to be. For example the ‘Pantone’ system attempts to build from a much larger palette of pigments than most printing systems (and still fails the inverse ‘metamerism’ problem – ie a Pantone strip can match a paint chip in the store but not on-site).

    But, as I point out at , actually I think it’s just wishful thinking to believe that colours are really well-defined properties of objects. It’s just light that has well-defined colours and the “colours” we assign to objects are the result of an unconscious mental process which seeks a consistent story about the various colours in different parts of our visual field. That such a story is not uniquely defined and is in fact dependent on the mind (and maybe past experience) of the observer is, I think, well illustrated by the striped dress conundrum that I referred to in response to your previous post.

    • Keith Allen

      The idea that only lights really have well defined colours is an interesting one, although I’m not sure why the colours of lights would be any better defined than the colours of objects? At any rate, I agree that there are unconscious mental processes that are involved in assigning colours to objects in our environment, and I agree that the way this happens may vary and depend on past experience, but I think this is consistent with holding that the visual system is trying to pick out mind-independent properties in the environment.

      • Light sources typically have a fixed spectrum which excites the three kinds of cones in our retina in a fixed proportion which might sensibly be defined as a raw apparent colour (and which is indeed a mind-independent property of the light source). I add the modifier “raw” because the the same raw apparent colour may be perceived by the observer as various different ultimate apparent colours depending on what other raw apparent colours are present in the visual field. And the algorithm for translating a raw apparent colour (seen in the context of being just part of a larger visual field) to an ultimate apparent colour in the mind of the observer seems to me to be inherently mind-dependent (and may even be influenced by the past experience of the observer), but the space of raw apparent colours depends only on the physiology of the observing eye.

        The light received from objects and pigments on the other hand depends on how they are illuminated, and so an object or pigment does not even have a well-defined raw apparent colour that is independent of how it is illuminated. Under any particular light source of course it does have a specific raw apparent colour, and the mental adjustment from raw apparent colour to ultimate apparent colour is an unconscious attempt to infer what the raw apparent colour would be under some average of normal sunlight conditions.

        This target (ie the raw apparent colour, or RGB value, reflected from “normal” sunlight) might be taken, I suppose, as your mind-independent real colour of the item. But since the quality of sunlight has quite a range, even this is not well-defined and a house which is seen as turquoise relative to its environment at one time of day may be seen as more silver-blue-grey at another. This would not contradict your thesis if all observers had the same reaction (we could just have a larger space of colours and call a pigment that acted this way something like “ocean-teal” for example). But in fact it is my experience that different people seem to have different reactions in these cases and so that the space of recognizable pigment colours is indeed mind-dependent.


Leave a Reply

Your email address will not be published. Required fields are marked *