Tuesday, November 24, 2009

Color naming follow-up

I received some interesting feedback on the color naming tech. report. A reader implied that in traditional psychophysics experiments the informants are first screened for color vision deficiencies (CVD), and our on-line experiment fails to do that. This was felt to be a grave omission especially in light of the two recent JOSA A papers by Kimberly Jameson and Natalia Komarova: http://www.opticsinfobase.org/josaa/abstract.cfm?URI=josaa-26-6-1414 and http://www.opticsinfobase.org/josaa/abstract.cfm?URI=josaa-26-6-1424.

The reader has a point, and there are simple on-line tests, usually based on the Farnsworth Munsell 100 Hue Test (FM100). However, we figured one of the advantages of crowd-sourcing and the feedback mechanisms built in our tools would be that "bad data" from anomalous trichromats and dichromats would just "disappear" statistically by getting a very low weight. By skipping a CVD test we can lower the bar of entry and get more informants.

The interesting aspect of the Jameson-Komarova papers are the tetrachromats. It can be speculated, that the FM100 scaling is incorrect for a tetrachromat, especially in the chartreuse region. If trichromats can only perceive a subset of what a tetrachromat perceives, then all uniform color spaces like CIELAB and CIECAM, which are scaled by screened trichromats, would have to be redefined by being scaled by tetrachromats. Would that imply that we have to replace tridimensional colorimetry with a new quadridimensional colorimetry, of which current CIE colorimetry would be a subspace?

In this light, there is no visual test to screen for tetrachromats. Although the genetic predisposition is well known, it would not help to ask female informants to submit us a DNA sample. In fact what really counts are not the single nucleotide substitutions at codons 180, 277, and 285, exon 3, but their expression, because the messenger RNA (mRNA) does not transcribe all genes. Today, the only way to identify tetrachromats from their genetic makeup would be to examine the cones in the retina, and we cannot ask informants to submit a piece of their retina.

The Jameson-Komarova papers study the development color categorization in a population. This is slightly different from what we are doing with the color naming experiment. Jameson-Komarova are studying the ontogeny of color categorization, while when we write about ephemera of color naming in our experiments, we mean the more short-term changes and vagary due to fashion and culture. We are more interested in operational models than in cognitive or physiological visual models.

Finally, we use the methodology of statistical surveys. The Jameson-Komarova papers adopt a completely different methodology: they model the visual system and then they use montecarlo methods to analyze the model statistically.

1 comment:

  1. Well, you received a good feedback, as your English colour naming database probably consists of more than 10% "bad data" of possible colour deficient participants. In the case of finding the centroids (averages) this percent may "dissapear" but in the case of finding the borders (distribution) of each colour category it may have a "neutral" appearance effect.