Many of the key issues in emotion research revolve around the interrelations of concept, word, and feeling. These interrelations are rarely examined in natural situations - they are either artificially modelled, with loss of validity, or naively assumed. This chapter begins with an outline of the ways in which emotions can be said to have meaning. Discussion of reference draws on Indonesian examples and leads into a consideration of Lutz’s Unnatural emotions, which presents a radically non-Western conception of emotion. Contrary to Lutz, I show that the practical import of emotion categories and clusterings cannot be predicted from tests or vernacular definitions; nor can it be derived from ‘discourse’. A sticking point for relativists, lack of a meta-category of ‘emotion’ and of presumed universal emotion concepts is regarded as a challenge to cross-cultural theory. How can we approach emotional phenomena for which we have no words, or grasp words for which we have no common reference? I look at Wierzbicka’s universalist project on emotion scripts, and draw on ethnographic examples of Chewong and Tahiti to tackle issues of hypocognition and radical alterity. The chapter closes with a narrative example that reveals the complexity of emotion-language use in natural situations.