The paper shows how the methodology for a pedagogical experiment can be improved through including the pre-research stage. If the experiment has the form of a test procedure, an improvement of methodology can be achieved using for example the methods of statistical and didactic analysis of tests which are traditionally used in other areas, i.e. measuring the learning outcomes. In particular, we deal with measuring the impact of different text styles of educational texts on the rate of knowledge transfer. For this purpose we are using the methods of statistical and didactic analysis. Statistical methods allow us to determine working hypotheses on statistical significance of differences in didactic characteristics of tests used for measuring the learning outcomes. The working hypotheses, whose validity have not been rejected, show us the weak points of the experiment and allow us to improve it before the experiment has been run in the full scale.