In many languages, emotion-denoting expressions contain body-part words. Words referring to abdominal organs (the heart, the stomach or belly, the liver) are particularly frequent in this context: for instance, in English "be broken-hearted", in French "avoir le coeur brisé", "avoir la peur au ventre". The semantic interpretation of body-part terms within emotion-denoting expression, and their figurative value in particular, are the subject of much discussion. Such expressions have regularly been treated as figurative "by default", but some authors have warned against these ill-considered figurative interpretations. Goddard (1994) and Enfield (2002), for instance, remind us that body-part terms in emotional expressions can also be polysemous, i.e. can denote both a body-part and an abstract seat of emotions (not identical to the body-part), or even denote an emotion as such (not a location at all). Interpreting bodypart terms as polysemous implies that the emotional senses of these forms are literal, and therefore that these body-part terms serve no figurative roles with respect to emotions. This article responds to Goddard and Enfield's call for more attentive analyses of the status of body parts in emotion-denoting expressions. I present a detailed study of the role of kangu "belly" in the emotional lexicon of Dalabon, a language of Northern Australia (of the Guniwnyguan family). I show that in this case, both the figurative and the polysemous (or literal) interpretations of kangu "belly" are justified by linguistic analysis, depending on the perspective adopted. If one considers emotional expressions with kangu "belly" as a set, they can be regarded as a system of coherent tropes, thereby serving a figurative function. On the other hand, if one compares expressions with kangu "belly" with Dalabon's emotional expressions featuring other body-part nouns, then the polysemous (literal) interpretation is justified. There seems to be no reason to discard either one or the other of these interpretations: both are analytically grounded, and speakers seem to endorse both of them in turn. This article explores the details and implications of each of these two analyses. Neither one is possible without the emotional lexicon being considered in its entirety, and until the semantics of emotional lexemes is thoroughly understood.