Abstract
Categorical perception (CP) is characterised as a superior ability to discriminate stimuli when they belong to different categories (cross-category) than when they are the members of the same category (within-category). Colour perception is well-documented to show this pattern. This property of colour perception has been used to test effects of language and learning, with some findings suggesting that colour CP may not be inherited but acquired as a result of learning the colour categories of a language (Özgen and Davies, 2002, JEP: General, 131(4), 477–493). Although such findings lend support to the well-known Linguistic Relativity Hypothesis, there is a serious problem in the literature. Virtually no study of colour CP has really tested perception of colour directly. Instead, tasks involving memory (e.g. successive same-different judgements with an ISI of variable duration) or sometimes subjective judgements of similarity have been commonly used to infer perception of colour.
We used a 2-Alternative-Temporal-Forced choice task where two hues to be discriminated were presented side-by-side simultaneously (without a spatial or temporal gap in between), making it an edge-detection task. In two experiments we investigated effects of category learning and language on colour CP using discrimination thresholds (with ZEST) as our measure. In Experiment 1 we trained observers on a novel category boundary (separating two kinds of green) and obtained threshold measurements before and after category learning. We found that following training, thresholds dropped considerably along the novel boundary (cross-category) but stayed unchanged elsewhere (within-category). In Experiment 2 we compared native English speakers with speakers of African languages which encode blue and green with a single term (no blue-green boundary). Africans had higher detection thresholds across the blue-green boundary than English speakers, while in neighbouring regions no such difference was found.