Children's reasoning and learning about levers and simple machines were investigated in this study. The study included several cycles of design, test and clinical interview tutoring sessions and the two final cycles are presented here. The methodology combined the use of qualitative clinical interviewing data and quantitative summative data: quantitative evaluations provided an overview of the lessons' effects, while qualitative, formative lesson evaluations allowed deeper insights into learning and reasoning processes. Three groups of participants were interviewed about the pretest, lesson and posttest. The pre- and posttests were standardized, and several new and widespread misconceptions about levers have been discovered that are less accurate or general than conventional conceptions. In experiment 1 the pre-posttest comparison between the control group and experimental group 1 showed that there were no differences and the instruction in experiment 2 was revised considerably as a result of the formative evaluation findings. Significant improvements were apparent for experimental group 2 with regard to conceptual change and for transfer when compared with experimental group 1--evident in group 2 students' ability to transfer their acquired knowledge to complex and compound levers and in conceptual changes apparent in simple levers questions. Lesson 1 was essentially a bridging lesson where "intuitive anchoring" examples were extended analogically via intermediate bridging cases to a target situation. The findings from lesson 1 suggested that reasoning from extreme case situations of levers might be instructionally useful, and this hypothesis was confirmed by results from experiment 2, where the instructional sequences based on extreme case reasoning proved to be powerful facilitators of the construction of mechanistic models by the students that fostered conceptual change and learning. The following directions for further research are suggested: students' conceptual models have implications for teaching and learning that are poorly understood at this stage, and research on instruction that employs experts' non-formal reasoning strategies should be encouraged.