Many languages of the world have connectives to express causal relations at the discourse level. Often, language users systematically prefer one lexical item (because) over another (even highly similar) one (since) to express a causal relationship. Such choices provide a window on speakers' cognitive categorizations, and have been modeled in previous work in terms of subjectivity. However, a broader empirical basis and a more specific operationalization of subjectivity are urgently needed. This paper provides in these needs by developing an integrative empirical approach to the analysis of the Dutch connectives omdat ‘because’ and want ‘since/for’ in written text, conversation, and chat interactions. These can be considered a case in point for linguistic categorization since related European languages show similar distinctions. The construct of subjectivity is decomposed into characteristics like type of relation and subject of consciousness (who can be considered responsible for the causality?). The use of statistical methods specifically suitable for hypothesis testing in natural language corpora produces results that provide new insights into the division of labor between the two connectives, as well as into the notion of subjectivity.