Scientific rhetoric can have a profound impact on the perception of research; it can also drive and direct further research efforts. What determines whether results are discussed in a neutral or a judgmental way? How precise and convincing must results be so that authors call for significant policy changes? These questions are in general difficult to answer, because rhetoric on the one hand, and content and methodology of the paper on the other, cannot be separated easily. We, therefore, use a unique example to examine this question empirically: the analysis of gender wage differentials. Here, the Blinder-Oaxaca decomposition represents a standard research method that compares male and female earnings, holding productivity constant. We analyze close to 200 papers to investigate what drives authors to talk about “discrimination”, whether and when they call for policy activism or when they are more hesitant to do so. Furthermore, we examine whether the rhetoric used really reveals an author's prejudice on the topic which may also be reflected in data selection and thereby his or her findings.