The concept of ‘identification’ remains a commonly called-upon resource for considering how media audiences might be influenced into taking up moral and cultural positions. Yet very little empirical evidence exists to support its claims; and recent critical conceptual work has significantly undermined many constituent parts of it. This article draws upon the very large data set gathered in the course of the Lord of the Rings international audience research project, to mount critical tests of the concept’s claims. The article then uses a combination of quantitative and qualitative evidence to explore the different bases on which audiences chose nine of the films’ characters as their favourites. An alternative approach to theorizing audience relations to characters is briefly outlined.