Abstract
The visible world is highly variable, containing surfaces with a wide range of reflectances (between 4% and 90% for charcoal and snow respectively). As we can discriminate hundreds of levels of lightness, describing such a variable world effectively would seem to require a large vocabulary. Despite this, the languages of technologically primitive cultures, which developed within largely natural environments, possess only two basic lightness terms, while those of more technologically advanced cultures, which inhabit more man-made environments, have only three (e.g. black, white and grey). Why should this be? Here we show, using information theory combined with estimates of; i) the distribution of reflectances within different visual environments, and ii) the uncertainty in recall of surface reflectance across subjects after a minutes-long interval, that in woodland environments only two lightness terms may be stable within a language, while in urban and indoor environments up to three terms may be stable. Hence, by formalising lightness description as the transmission of lightness information from the environment (variation in surface reflectance) in the presence of noise (variation in recall of reflectance) we are able to produce veridical estimates of the number of lightness terms which may be stable within a language within a given visual environment. Determining why a language has the colour terms it does can provide valuable insight into the philosophical issues surrounding the relationship between language and our internal representation of the world. Traditionally, the linguistic partitioning of colour space (and its lightness dimension) is viewed as either; i) learned and arbitrary (Relativists), or ii) the result of ecological or physiological constraints (Universalists). Our results suggest that our basic lightness terms are an optimisation, given our cognitive limitations, to the reflectance characteristics of the visual environment we inhabit.