Abstract
Categorical colour perception could originate from either physiological (universal) or linguistic (relativist) constraints. Both positions assume perceptual and linguistic colour categories to be isomorphic yet a flexible relationship of mappings may exist from which different categorical organisation could arise.
We tested this hypothesis by asking 40 English-speakers with normal colour vision to sort or name 140 Munsell chips (20 hues each at 7 values) over three tasks: 1) a Free Sorting Task (FST) with an unlimited number of categories; 2) a Constrained Sorting Task (CST) where the number of categories was limited to 8; and 3) a Constrained Naming Task (CNT) using 8 basic colour terms.
Data was analysed using cluster analysis and multidimensional scaling to identify colour categories and to derive a two-dimensional model of colour representation. The grouping of 68.6% of the samples obtained for 8 clusters is identical across tasks. The main difference was accounted for by the existence of a luminance-based pastel category (grouping of value 9 chips) in the CST, and a turquoise hue category in the FST replaced by the orange hue category that was present in the CNT.
In addition, representation of the 20 hues (averaged across the 7 values), draw the Munsell hue circle along a Red/Green and Yellow/Blue dimension. Large inter-points distances corresponding to category boundaries varied across tasks, with the exception of the largest common distance (5PB-10PB). Finally, the estimated collective consensus of category organisation was 69.2% for the FST, 73.8% for the CST and 85.4% for the CNT.
Structure of the colour space is independent of the task, but inter point distance variations generate differences in colour category mappings. Furthermore, constraints imposed by colour naming produce the strongest collective consensus to fulfil successful communication.