Whorfian effects on colour memory are not reliable

Abstract

The Whorfian hypothesis suggests that differences between languages cause differences in cognitive processes. Support for this idea comes from studies that find that patterns of colour memory errors made by speakers of different languages align with differences in colour lexicons. The current study provides a large-scale investigation of the relationship between colour language and colour memory, adopting a cross-linguistic and developmental approach. Colour memory on a delayed matching-to-sample (XAB) task was investigated in 2 language groups with differing colour lexicons, for 3 developmental stages and 2 regions of colour space. Analyses used a Bayesian technique to provide simultaneous assessment of two competing hypotheses (H1–Whorfian effect present, H0–Whorfian effect absent). Results of the analyses consistently favoured H0. The findings suggest that Whorfian effects on colour memory are not reliable and that the importance of such effects should not be overestimated.

Keywords:

• Whorfian effects
• Linguistic relativity
• Colour categorization
• Colour memory

We wish to dedicate this article to Kemuu Jakurama, our guide and interpreter in Namibia, who passed away in 2012. We are also grateful to several reviewers and the journal editor for the valuable feedback they provided.

We are grateful for the support of an Economic and Social Research Council Research Studentship to O.W. and a British Psychological Society study visit grant to O.W. Manuscript preparation was supported by a European Research Council Starter Grant [project CATEGORIES, grant number 283605] to A.F.

Notes

1 The illumination provided by the Macbeth lamp used to test English participants was considerably lower than the daylight in which Himba participants was tested, making the task relatively more difficult for English than Himba participants. However, color temperatures were similar. Consequently whilst differences in illumination may have influenced overall error rates, patterns of errors in the groups tested are unlikely to have been influenced by such differences. Goldstein et al. (2009, pp. 231–233), for instance, compared performance of two groups of Himba toddlers under conditions of artificial versus natural lighting and found that differences in illumination had no effect on performance.

2 The direct labelling theory predicts that participants who categorize all three stimuli in a set using the same basic color terms should make more XAB errors than participants who categorize stimuli using different terms. However, labelling theory also predicts that aaa and abc of categorization should produce equivalent error patterns on 1A and A2 trials.

3 In NHST, p-values do not represent the probability that a null hypothesis is true. Rather, they represent the probability of the observations/data if the null hypothesis were true.

4 When the analysis of the XAB task was repeated excluding data from participants who performed at floor and ceiling, the outcome was not substantially different.