Statistical Power for Small Effect Sizes: An investigation of backward priming in Mandarin-English bilinguals

Date

2024-01-29

Authors

Li, Xiao Xiao

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Backward priming, or L2 to L1 priming, is a small but important effect for understanding the structure of the bilingual lexicon. A meta-analysis of priming in bilingual populations has shown that while the backward priming effect is quite small, it is qualitatively but not quantitatively different from the forward (L1 to L2) priming effect (Wen & Van Heuven, 2017). The empirical evidence for this view has come from various groups of bilinguals, including Japanese-English (Nakayama et al., 2016) and Korean-English (Lee et al., 2018) bilinguals, but not yet with Mandarin-English bilinguals: In this population, the effect is inconsistently significant. In response to this, researchers have raised the question of whether the existing studies were underpowered, given the small backward priming effect. Using a simulation-based power analysis, I show that this is most likely the case, as roughly 5400 observations per condition are necessary to detect a small backward priming effect. Previous work collected an average of 453 observations per condition, making it very unlikely for their statistical tools to be able to detect the effect. Based on this, I recommend that future work in this field conduct power analyses a priori, using the results as a guideline rather than a strict criterion for adequate power. Adopting this practice can help make experiments more replicable and future work in this direction is crucial for developing our understanding of the structure of the mental lexicon.

Description

Keywords

psycholinguistics, statistical power, bilingualism, experimental methods, bilingual lexicon

Citation