In recent years, anthropology and cultural evolution have increasingly used text mining techniques to explore long-term cultural dynamics across decades and centuries, analysing large datasets of written documents. This talk focuses on findings that reveal a growing prevalence of negative emotional content, particularly in English-language literature and music. Using this case study, I will discuss key methodological challenges faced by this line of research, including ensuring the representativity and validity of datasets, formulating clear, theory-driven, questions, and addressing issues of causality.