In this paper, a corpus-based approach is used in order to provide an in-depth analysis of two German emotion words (Wut and Zorn) that roughly correspond to the English lexical unit anger, establishing their similarities and differences on different levels of meaning. For the corpus study, data from two very large corpora of contemporary German texts (fiction, non-fiction, scientific and newspaper) are used, DWDS (Digitales Wörterbuch der deutschen Sprache des 20. Jhs.) and the collocation database CCDB, through which collocation profiles extracted from the German Reference Corpus (DeReKo). As a first step, the metaphorical and metonymical conceptualizations of both items are analysed and compared. Other aspects that are relevant for the description of these emotions and can be described by means of co-occurrence analysis are their conceptual proximity to other emotion words and the way they are described or evaluated (their semantic prosody). In a third step, looking at the semantic subsets the co-occurrences belong to from the point of view of syntagmatic relations between the emotion concept and its surroundings in the text (semantic preference), information can be extracted about causes of anger, about its consequences and about who experiences it. The results show that, taking both words together, the metaphorical conceptualization of the concept of anger proves to be very similar to that which has been described for other Western languages like English or Spanish. Going into more detail, however, the corpus analysis allows us to explore how each word specializes in certain aspects of the emotion and to corroborate, contradict or complement previous studies on this.