The planning of adequate control groups is a central aspect of educational research. Didactical studies, however, are often field studies encountering many problems of everyday classroom teaching? Recent research has indicated that teachers and their beliefs have an enormous influence on learning and retention. Taking this line of evidence further, a devil's advocate may emphasize that differences in treatment-control designs may be not a result of the different instructional strategies but reflect differences between different teachers. In this study, we try to sort out this kind of "teacher effect" by comparing two approaches of a complex treatment-control group design. In the first approach we compare treatment and control groups that were taught by the same teachers (labeled "matched-pair tandem design"), and in the second approach we compare the control group with an unrelated treatment group where different teachers taught the treatment groups ("unrelated design"). When comparing the "matched-pair tandem design" with the quasi-experimental approach, we found i) similar patterns in both educational experiments, and ii) higher effect sizes in the unrelated, quasi-experimental design. (Contains 2 tables and 1 figure.)