The supervision offered at universities today is to some degree being questioned, while the situation of doctoral students is undergoing changes. The supervision students get nowadays is very different from what was available when today’s supervisors were doctoral students. The present study relies on thirty-one in-depth interviews with experienced research supervisors from all faculties at the one university. The Swedish governing documents regarding supervision have mostly been phrased in very general terms on the assumption that certain disciplines have several traits in common. Other documents describe the requirements for supervisors more sweepingly; everybody is expected to work according to the same lines. There is no question that supervision could be anything but good and no attempt at defining the task is made. Previous studies show that supervisors are unaware of what is really expected of them. The aim of this thesis is partly to investigate whether there are any communal traits in the accounts, at individual or group level, and partly to demonstrate how research supervisors’ views on research supervision can vary. The objective is also to highlight qualitative differences in the descriptions by the supervisors as to how and why they supervise. Results show that it is not easy to articulate knowledge about ones own actions. The supervisors under study have not previously reflected too much on the question of supervising. They have neither received nor requested feedback and do not expect to get honest verdicts from their students. They profess themselves to have developed a mode of supervision on their own, without assistance or any form of training. The thesis is based on a theory of variation, tacit knowledge and reflection and that supervisors will be shaped by connections to certain Communities of practice. It draws on the silent or unarticulated knowledge of what supervision is about, how and why supervision is carried out in a specific way, at group level or individually. At group level some similarities to other investigations can be found. Traditions regarding the form of the thesis are usually compliant with faculty norms. At the individual level, however, the study presents new findings in pointing out the existence of three different styles of supervision: researcher, leader and official. These can not be traced back to any specific disciplines. The three styles differ from each other mainly in their attitude to the doctoral students and in their outlook on the question of power and responsibility. The most important conclusion is to draw attention to the significance of having an individual perspective. In order to be able to improve the art of supervision, getting interaction and feedback from doctoral students and the organisation is not sufficient. A prerequisite for development is to gain an awareness of ones own actions. In order to become conscious about ones doings it is necessary to give and take feedback about oneself, achieved through self-reflection.