Mirjam, randomized controlled and controlled studies are difficult to design in medical education because such design will affect the schedule and the teaching and learning meant to be used. Such studies also do not attract higher citations as it is the case in science and medicine. (see for example my analysis: https://www.researchgate.net/publication/43100386_Research_in_medical_education_is_not_just_on_telling_a_story). However, when you compare two teaching and learning methods under the same conditions, you can clearly demonstrate the impact of the intervention. From my experience, I used two methods in teaching surface anatomy (published in Medical Teacher), and I conducted the study during sessions committed to teaching students learning skills to avoid the challenge I mentioned earlier about difficulties in designing controlled studies. There are other studies I remember reviewing some to Anatomical Science Education, BMC Medical Education, and Medical Teacher.

It is possible, instead of comparing teaching and learning methods by pre and post-test instruments, by postulating a cognitive target (hypothesis) that might be reached by a teaching and learning sequence designed in terms of models.

This reference might help you if you read Spanish or can translate it: López-Mota, A. & Moreno-Arcuri, G. (2014). Sustentación teórica y descripción metodológica del proceso de obtención de criterios de diseño y validación para secuencias didácticas basadas en modelos: El caso del fenómeno de la fermentación. Revista Bio-grafía, 7(13), 109-126. 

I suggest you look into quasi-experimental designs. Classical texts are by Donald Campbell, William Shadish and others.

At the University of Salzburg, Jean-Luc Patry & team used this design for their Value and Knowledge Education Program VaKE, using a constructivist approach.

Also useful: An Applied Reference Guide to Research Designs (Edmonds/Kennedy, 2013). I am quoting from Part II Quantitative Methods for Nonexperimental Research: ".. nonexperimental research is utilized when the variables of interest cannot be controlled through the means of manipulation, inclusion, exclusion, or group assignment" (p. 96)

Chapter 11 deals with Grounded Theory Approach (see Glaser/Strauss, 1967), which may also be relevant.

You could also have a look to: Science Education Research: López-Mota, A., López-Valentín, D., Rodríguez-Pineda, D. & Gutiérrez, R. USE OF MODELS IN DESIGNING AND VALIDATING TLS: AN EXAMPLE FROM CHEMISTRY, in Engaging Learners for a Sustainable Future (Proceedings of ESERA 2015) ISBN 978-951-51-1541-6

There are a number of well-designed RCTs in education - I think that you will find the refs in

Using Random Allocation to Evaluate Social Interventions: Three Recent U.K. Examples

Ann Oakley, Vicki Strange, Tami Toroyan, Meg Wiggins, Ian Roberts and Judith Stephenson The ANNALS of the American Academy of Political and Social Science 2003; 589; 170

Esteemed colleagues,

thank you very much for these answers, which have proven very helpfull allready. The references are of very good use, and it is also very interesting to see the different diciplinary perspective taken. Personally, I find it interesting that more open-minded writings/approaches to field studies as a valid research design seem to come from discipline other than the pedagogical-psychological sphere, which relates to different paradigms, I guess.

However, as mentioned in the answer by Samy Azer, the most crucial point for a valid field-study research desing still is how to compare your construct of interest in goups under the same condition(s). That, in my experience, is also the hardest part to realize, since in a naturalistic environment conditions are nealy never exactly the same, which makes the design always fragile....

Would love to read your comments on that!