We report on our experience in teaching introductory courses on programming based on formal specification and program calculation, in two different Computer Science programmes. We favour the use of logic as a tool, the notion of program as a formal entity, as well as some issues associated with efficiency. We also review and use in practical cases some program transformation strategies, such as generalisation, tupling and modularisation. We describe our approach, its advantages and drawbacks. Furthermore, we present some preliminary results from an ongoing qualitative research which intends to characterise, describe and understand the students' experiences when taking these courses.