Abstract
Constrained color-sorting—the partition of a color palette into sequential numbers, n, of categories or “piles,” n = {2, 3, …, N}—has been used to examine universal vs. culture-dependent processes that underlie color-term evolution. Here we use sorting to test two hypotheses related to these ideas. 1) If the constrained color sorts (CCSs) depend solely on universal processes that guide color-term evolution, then CCSs should be relatively insensitive to the exact composition of the color palette. 2) Otherwise, CCSs should at least be “optimal” (as measured by well-formedness: Regier et al., 2007). We report three studies, each using a different palette, where English speakers sorted the color samples into 2 … 6 piles; one study also included Somali speakers. All palettes included good examples of the Basic Color Categories (BCCs), but they differed in how hue, saturation, and lightness covaried. Results showed that pile-sorts reflected psychologically salient features of colors, which were mostly linked to BCCs. However, 1) the unfolding of sort-categories with increasing n differed greatly across palettes, and the order in which the sort-categories appeared always deviated from predictions based on classical theories of color term evolution; 2) pile-sorts were not optimal. Taken together, our results reveal strong dependence of CCSs on pre-existing lexical categories but only weak dependence on universal pre-linguistic representations of color related to color-term evolution.
Funding: Funding: National Science Foundation BCS-1152841 to DTL