Instead of answering this question directly, I want to start with another question, one that motivates asking the question of my title. How do humans first come to know about the categorical colour properties of things, properties like redness and blueness? Start with the simplest possible story. Red things visually appear one way to you, namely as red, and blue things visually appear another way, namely as blue. As Byrne and Hilbert put it:

‘If someone with normal color vision looks at a tomato in good light, the tomato will appear to have a distinctive property—a property that strawberries and cherries also appear to have, and which we call ‘red’ in English’ (Byrne and Hilbert 2003, 4).

These facts about visual appearance enable you to lock onto colour properties like red in thought. This in turn enables you to come to know that certain objects are red. Fodor relies on a story along these lines. He writes:

‘all that’s required for us to get locked to redness is that red things should reliably seem to us as they do, in fact, reliably seem to the visually unimpaired’ (Fodor 1998, 142).

This is perhaps the simplest story it is possible to tell about the developmental emergence of knowledge of colour. But is it a true story?

First objection: development

All the available evidence suggests that the story is wrong in two ways. First, the claim that we get ‘locked to redness’ (are able to think about red things as red) by virtue of red things appearing to us as they do is likely to get things back-to-front. It turns out, counterintuitively, that whether red things strike us in any particular way probably depends on our being ‘locked to redness’.

Part of the evidence for this claim comes from development. Let me start with some background. From around four months of age or earlier, infants are able to visually discriminate some categorical colour properties. Some of the best evidence for this reveals that categorical colour properties can produce pop-out (Franklin, Pilling, and Davies 2005) and oddball effects (A. Clifford et al. 2009) in infants. But whereas infants in their first months of life do not comprehend or produce colour words (or indeed any words), which categorical colour properties adults can visually discriminate depends on which colour words they use (Kay and Regier 2006; Roberson and Hanley 2007; Winawer et al. 2007; Thierry et al. 2009), and colour words can vary dramatically between cultures (Roberson, Davies, and Davidoff 2000). This indicates that the categorical colour properties infants can visually discriminate are not generally the categorical colour properties that they go on to name or acquire concepts of later in life. For instance, infants appear to discriminate blue and green in a way that not all adults do (Franklin, Skelton, and Catchpole 2014).

How do infants eventually become able to visually discriminate approximately the categorical colour properties that the adults around them visually discriminate? Anna Franklin and her colleagues studied children in the process of learning to use their first colour words accurately. They measured how accurately children of different ages used colour words and how accurately they could visually discriminate the categorical colour properties for which their language has words. Surprisingly, they found that more accurate use of colour words was associated with weaker abilities to visually distinguish categorical colour properties corresponding to words (Franklin et al. 2005). This suggests that, for children learning their first colour words, being able to visually discriminate the categorical colour properties named by words is not a cause but a consequence of mastering colour words. Starting to use colour words initially makes you worse at visually discriminating the categorical colour properties those words name because, at this point, the categorical colour properties you can visually discriminate are distinct from the categorical colour properties your words name. The shift from visually discriminating infant categorical colour properties to visually discriminating categorical colour properties for which there are words in English, Himba or Russian is a consequence of learning a system of colour words.

This view is indirectly supported by discoveries about adults. In adults, the ability to visually discriminate properties denoted by particular colour words like ‘red’ depends not only on having learned to use words for those colours accurately in the past (Ozgen and Davies 2002; Winawer et al. 2007; Zhou et al. 2010) but also on being able to activate some component of the ability to apply the colour word at the time a stimulus is presented (Roberson and Davidoff 2000; Pilling et al. 2003; J. A. Wiggett and Davies 2008). This makes it less surprising that infants’ learning to use colour words accurately is a cause, not a consequence, of their abilities to visually discriminate the corresponding (adult) categorical colour properties.

So far our focus has been on words and visual discrimination, whereas the view I want to challenge is about knowledge and experience. But maybe these are linked. Consider two claims. First, being unable to visually discriminate a categorical colour property, even in ideal circumstances, is a symptom of things not visually appearing to you to have that property. Second, those who can use a colour word accurately are ‘locked to’ (as Fodor puts it), and in a position to know facts about, the corresponding categorical colour property. I’m not certain these claims are both true, so let me be cautious. If we take the evidence at face value, we must either reject at least one of these claims or else accept that red things appearing to us in whatever way they do is a consequence, not a cause, of our being locked to redness.

Second objection: phenomenology

Earlier I wrote that the simplest possible story about the developmental emergence of knowledge of colour is wrong in two ways. So far I’ve only described one: locking on to a categorical colour property is more likely to be a cause than a consequence of the difference it makes to your experience. What else is wrong with the simplest possible story?

Red things do not visually appear red, or so all the available evidence indicates.

Before I get to the evidence, let me sketch the key idea behind several studies. Two things that vary in categorical colour property will, of course, also vary in which particular shade of colour they are. So the fact that two such things have different visual appearances might be due just to the difference in their shades and not at all to the difference in categorical colour property. How can we distinguish differences in appearance due to shade from differences in appearance due to differences in categorical colour property? Fortunately it is possible to find sequences of shades of colour where any pair of adjacent shades are as perceptually similar to each other as any other pair of adjacent shades. This allows us to compare a change in shade which does not involve a change in any categorical colour property of interest (two blues, say) with a comparable change in shade which does involve a change in a categorical colour property of interest (a blue and a green, say).

Which categorical colour properties are of interest depends on the individual subject. For a typical adult, there will be a set of categorical colour properties for which she has colour words and which are such that she is faster and more accurate in distinguishing things when they differ with respect to one of these categorical colour properties than when they do not so differ (faster: Bornstein and Korda (1984) ; more accurate: Roberson, Davidoff, and Braisby (1999), p. 22-7 ; see also Witzel (2014)). These differences in speed and accuracy are unlikely to be artefacts of failure to correctly measure the perceptual similarity of particular shades of colour because much the same differences are observed for newly trained categorical colour words (Ozgen and Davies 2002; Zhou et al. 2010; A. Clifford et al. 2012). Further, discrimination of these categorical colour properties involves pre-attentive processes which are automatic in some of the senses that perceptual processes are (Daoutis, Pilling, and Davies 2006, A. Clifford et al. (2010)). An adult will have visual experiences of these categorical colour properties if she has visual experiences of any categorical colour properties.

How can we tell whether differences in any categorical colour properties give rise to differences in visual appearances? A difference in visual appearance typically has several consequences. Typically, the more things differ in visual appearance, the less likely someone is to judge that they are similar. And typically, the more similar things are in visual appearance, the more likely they are to be visually grouped together. But differences in categorical colour property seem to not affect judgements of similarity (Davidoff et al. 2012,witzel2014category). And they also seem not to affect perceptual grouping (Webster and Kay 2012). This is evidence that red things do not visually appear in any particular way.

What can we conclude? I see three possibilities. We could withhold judgment on the basis that there are not many studies yet (although in that case we should note neuroscientific evidence indicating that categorical colour properties are processed awfully late: He et al. (2014)). Or we could insist that categorical colour properties like red do have visual appearances, albeit ones which are too subtle to detect by ordinary means and ones for whose existence there is no evidence. Or we could reject the view that there are visual experiences of red and other categorical colour properties.

Note that this last possibility does not require holding, implausibly, that differences in categorical colour properties make no difference at all to the phenomenal characters of experiences. Differences in categorical colour property surely do make a difference to the overall phenomenal character of experiences. But not all phenomenology is a matter of perceptual appearances (or so I’ll be suggesting in my next post).

Conclusion

The simplest possible story about the developmental emergence of knowledge of colour conflicts with the available evidence on two counts. The phenomenal effects of categorical colour properties are more likely a consequence than a cause of knowledge of categorical colour properties, and there are most likely no visual appearances of red or other categorical colour properties.

This is why we need an different story about how humans first come to know about the categorical colour properties of things. The evidence we’ve been considering suggests that the true story will involve pre-linguistic visual discrimination plus learning colour words through communicative interactions.

References