Hi Ed. It was interesting to read your comments on the natural language technology I’ve been working on for the past year or so. You raise a really interesting point that really got me thinking. Unfortunately, after thinking it through, it seems to me that we cannot use this technology to gain insight into how different people perceive colors. The reason for this is that color perception lies entirely with an individual’s consciousness. Color naming, however, is a learned response. We learn, generally in our early childhood, to associate a certain name with the perception that results from observing a particular colored object. So, while virtually everybody learns to associate the name ‘blue’ with the color of the sky on a clear, sunny day, we cannot be certain that the internal perception of that color is identical for each person.
We know that the physiological make-up of the human visual system can vary markedly from one individual to another. It is fair to assume that this results in the signals being sent from the visual system to the brain to differ between individuals. Still there is remarkable agreement in the color naming behavior across a range of individuals. This is simply because we have all learned to associate the color name ‘blue’ with the color perception that results from looking at the sky.

Response to David’s comment:
Hi David. The issues you raise are certainly important considerations. In fact similar questions were asked by several of the scientists at the Inter-Society Color Council conference where I recently presented my results. As it turns out, the examples you give regarding arguments over color names do not affect this technology significantly. There are many synonyms for a single color stimulus and the technology is able to map multiple names to the same region in color space. A good example of this is ‘bluish-green’. This color might be called ‘cyan’ by some people and ‘turquoise’ or ‘aqua’ by others. The technology will map all these names to a similar region of color space. Another point to keep in mind is that color naming is not intended to be a precise and specific specification of color, so color names map to a fuzzy region in color space, not to an exact, specific point. There are certain color applications where this fuzziness is actually an advantage rather than a disadvantage. Color adjustment of images is one such example. By using fuzzy descriptors like color names we obtain results that are smooth and pleasing. Highly precise color specifications in such applications can lead to undesirable artifacts such as contours.

This post got me thinking about whether this tech could be used to study difference in how people perceive colours.

Since colour exists in the brain, one person’s idea of what ‘brick red’ looks like may be very different to another’s. ‘Brick red’ is most likely to apply to a wide range of combinations of light wavelengths, and it would be interesting to see how these ranges vary between individuals, throughout development, between ethnic groups etc.

After all, who hasn’t become embroiled in an argument about whether something is ‘actually’ blue or green, and so on?