Text mining methods applied in social sciences research are based on the premise that the analysis of texts leads to valid conclusions about social reality. The paper examines the theoretical and empirical aspects of this premise: first, it shows by recourse to the tradition of hermeneutics and the sociology of knowledge, how the relationship between text and social reality can be determined. Subsequently, it illustrates the heuristic performance of various text mining methods, using the bigram ‚social market economy‘ (ger. ‚Soziale Marktwirtschaft‘) as an example. The conclusion leads to the important analytical finding that only the constant procedural entanglement of Close and Distant Reading, referred to, in accordance with Moretti’s terminology, as Blended Reading, will allow the researcher to produce viable knowledge.