The educational potential of non-linear systems is given surprisingly little attention in science education research--at least in research that links content matter and educational issues. The project on educational reconstruction of non-linear systems at the IPN has investigated the educational significance of threferring phenomena and the effects of teaching and learning processes. In a nutshell, it has been investigated whether it is worthwhile to teach non-linear systems and whether students understand the basic ideas of these phenomena. In addition, a number of instructional modules addressing various characteristics of these systems have been developed. The work has been based on a model of educational reconstruction that, besides other issues, closely links analytical and empirical educational research with development of teaching and learning sequences. In the present paper a brief overview of the sketched project will be provided. Main emphasis, however, will be given to the teaching experiment, which has proven to be a valuable research method to investigate teaching and learning processes. Teaching experiments may be viewed as Piagetian critical interviews that are deliberately employed as teaching and learning situations. The interviewer assumes the roles of a 'classical' interviewer (who tries to understand students' understanding) and a teacher. Although the teaching experiment is a laboratory situation, it also shares major features of research in actual classrooms. Therefore, this method appears to be well suited for linking research and development in the first steps of designing teaching and learning sequences.