In the first phase of this study, a cognitive conflict strategy was used to design teaching/learning activities aimed at developing aspects of the newtonian concept of force. The effectiveness of the activities was studied in classroom research that took place in preuniversity courses in Botswana and South Africa. Pre-/posttesting showed a considerable increase in correct answers, but the answer-patterns were not consistent with the assumption that most students based their reasoning either on correct or on alternative concepts. Aspects of the initial premises of the conceptual replacement approach were questioned: In what way are student ideas "alternative," and what is the students' cognitive basis for their construction of scientifically better ideas? We will show that many of the students' beliefs about motion and its causes, often expressed as "motion implies a force," do not contradict scientific beliefs, provided that we accept that students, when they use the word "force," refer to a concept which differs from the scientist's concept of force. Students do not distinguish concepts as precisely as scientists, have beliefs that may be incorrectly generalized to unfamiliar contexts, and frequently express their beliefs in nonscientific terms. However, in our interpretation, students do not have beliefs about familiar situations that are incompatible with scientific beliefs. Conceptual replacement, therefore, is not an adequate strategy to foster conceptual growth for the topics under consideration. Rather, the students' prior correct beliefs need to be identified as a potential basis for development of the scientific concept of force. The revised interpretation of students' conceptions guided a revision of the teaching sequence in the second phase of this study, aimed at helping students to develop their own conceptual tools to perceive and potentially resolve dissonance before they experience it. The revised sequence, which is based on concept refinement and context expansion, resulted in increases in correct answers and answer-patterns more consistent with concept-based reasoning.