Conservation agriculture (CA) is based on 3 principles, namely reduced soil disturbance, permanent soil cover, and more complex and legume-rich rotations; and multiple studies have shown its positive impacts. Because CA relies on a variety of ecological processes, it is more deeply rooted in a specific ecological context than conventional agriculture. The complexity of these processes makes it difficult to elaborate general recipes to be applied by farmers, who therefore need to learn to make their own choices adapted to their own agroecosystem. Consequently, helping farmers to move toward CA requires supporting them in learning to develop their own practices. Farmers’ learning remains poorly investigated at the individual level, with in particular very little work focusing on learning in CA. We hypothesize that the processes involved in learning to practice CA may differ from those involved in conventional agriculture: for instance, the current lack of detailed reference documents may induce farmers to experiment more. Against this background, we here aimed at describing how farmers experienced in CA learn, by qualifying their learning mechanisms and processes. To do so, we conducted five comprehensive interviews with farmers experienced in CA, and then inductively analyzed the data to explore the diversity of learning mechanisms involved, i.e. the elementary actions or cognitive activities which, organized together, constitute a learning process. We, thus, propose a descriptive framework of non-ordered and non-obligatory learning mechanisms that appear to be mobilized by farmers experienced in conservation agriculture, as a first step toward a deeper analysis of their learning processes. We further emphasize the often unintentional aspect of learning, as well as the importance, for farmers who wish to implement CA practices, of developing new standards of comparison. A better understanding of these learning processes would help improving extension services and training for farmers.