We study potential drivers of the link between gendered languages and gender differences in economic behaviour that economists have recently documented. Combining identity economics and linguistic theory, our formal model distinguishes a direct effect of speaking a gendered language from indirect effects through gender norms and highlights pitfalls in discriminating these effects empirically, particularly when studying behaviour of immigrants. Our empirical exercises illustrate the severity of the problem, as self-selection and intergenerational transmission of traits appear to bias estimates of the link between migrant behaviour and gendered language.