Rolling off the Tongue into the Top-of-the-Head: Explaining Language Effects on Public Opinion

Growing evidence shows that mass opinion varies by interview language, yet modest theory exists to explain this result. I propose a framework where language impacts survey response by making some political concepts more mentally accessible. I claim that concepts vary by how associated they are with certain languages, which means people are more likely to acquire a construct when it is tied to the tongue one speaks. Hence, recalling concepts from memory should be easier when the language a construct is linked to matches the tongue one interviews in, thereby intensifying people’s opinions. I test my theory by manipulating the interview language in two U.S. surveys of English/Spanish bilingual Latino adults. I generally find that language influences the accessibility of concepts. For example, subjects report higher opinion levels for concepts that are tied more to their interview language, such as American identity among English interviewees. Subjects who interview in English are also less likely to refuse completing items measuring knowledge about U.S. politics, and more likely to answer them quickly. Items reflecting constructs that are highly labile (e.g. anti-Obama affect) or very crystallized (e.g., partisanship) do not display these patterns. I then rule out that language effects are mostly mediated by a heightened sense of anxiety, anger, pride or efficacy that emerges when bilingual subjects interview in one of their languages.