A first part of this paper gives a rough picture of some difficulties encountered in research, in education, and in problem solving, for integrating them to one another. One can notice a much too global characterization of cognitive processes and a lack in the characterization of semiotic aspects. A second part analyses some theoretical limits to this integration. They are mainly due to the current conception of memories unable to take into consideration the micro-cognitive-processes at work under the reorganizations of knowledge when actualized within the situation. A third part presents a way toward the integration research-education-problem solving, relying on a cognitive approach of Culioli’s enunciative theory of language, and presents some of the author’s data. Micro-cognitive-processes are depicted in terms of the construction of aggregates (declarative versus procedural ones, standing at different levels of internalization and externalization), and of different processes of detachment from the situation. Then several kinds of interactions allow an on-line identification of the constraints of the task. The characterization of these constraints seems basic for each of the considered areas, research, education, and problem solving.