Rethinking the relationship between pronoun-drop and individualism with Bayesian multilevel models

Can the language we speak determine how we represent the world around us? To those familiar with the theory of linguistic relativity, this may seem like an age-old question about which everyone has their own answer. Although the evidence supporting linguistic relativity remains controversial, the long reach of language into our perception and behavior is nevertheless an intriguing possibility that deserves further investigation. Here I take a closer look at a case of linguistic relativity that had a particularly strong impact on cross-cultural research: the pronoun-drop effect. The theory of pronoun-drop effect posits that languages that allow their speakers to drop subject pronouns in verbal communication would lead their speakers to create collectivistic culture. It was argued that the absence of pronouns necessitates the speakers to embed their self-identities in the context of social interaction, so the linguistic practice of omitting pronouns would reduce the sense of individuality in the minds of speakers. After conducting a series of Bayesian multilevel analyses on the original dataset, however, the current study concludes that the pronoun-drop effect is unlikely to be a robust, universal phenomenon. The analyses revealed that the majority of statistical signal supporting the phenomenon comes from the Indo-European language family, and other families provided little or inconsistent evidence. It was also observed that the Indo-European languages alone made up 61 per cent of the original dataset, and dropping them from analysis completely nullified the pronoun-drop effect. These observations suggest that the pronoun-drop effect is a consequence of failing to account for (i) varying effects among language families and (ii) overrepresentation of the Indo-European languages. With these results, this article suggests that the theory of pronoun-drop effect should be thoroughly revised. Additionally, the article provides several suggestions for many similar cross-cultural studies that suffer from the same problems as the pronoun-drop effect study.