Qualitative research is usually underestimated as a strategy to construct valid knowledge because its lack of perfect procedural rationality that is typically assumed to characterize the experimental designs. However, there are many educational situations in which qualitative research is particularly relevant. In this paper, the rationality of designs in qualitative research is discussed. By extending Rawls' distinctions among types of procedures (Rawls, 1971), Elgin (1996) proposes three models of rationality to classify epistemological theories: whereas the perfect procedural rationality justifies the experimental design and the pure procedural is connected to social constructivism, the imperfect procedural rationality underlies the system of highly conjectural design developed in the qualitative research. The experimental design implements a perfect procedural rationality because it holds strong commitments to standardization and formalization, its procedures are constraining, its results about causal relationships between variables are conclusive, whereas fidelity in design is respected (Eisner and Peshkin, 1990). Although from the point of view of the perfect procedural rationality qualitative researches are considered defective, they are relevant for inquiries in which the complexity of specific contexts and the diversity of the learning processes are crucial aspects to be highlighted. Qualitative researchers (see for example Bronfenbrenner, 1979; Cole, 1995; Saxe, 1991) propose different frameworks to analyse the relations between a complex system of educational contexts and the individual development; convergent findings make evident that learning contexts are multifarious and unlikely to be grasped by normative models and controlled research strategies; the reality of the developing subject cannot be reconducible to any unitary structure, being composed by etherogeneous cognitive schemes, in connection to the problems and the cultural resources offered. To inquiry the complexity of the educational processes, qualitative researchers look for contextualized data, and integrate three different epistemic actions: gathering information, focusing questions, analysing data (Lofland and Lofland, 1995). They work with a fairly loose and flexible network of epistemic commitments, each subject to revision or revocation, should defects emerge or improvements be found. Qualitative researchers can produce convincing and empirically stringent results, but they cannot reach conclusive ones, because they do not hold a set of well-established rules to guide inferential processes. Therefore, the rational justifications of qualitative research tend to be elusive because the strategies are rather idiosyncratic and lacking in canons and conventions. An imperfect procedural rationality can ground an independent and solid epistemological status for qualitative researches because it recognizes the merits of reflective equilibrium as a process of balancing competing claims against one another to produce convincing reasons, recognizes an independent evaluation criterion to assess the validity of each research, based on the public inspection of any single design, meanwhile it accepts that no well-established and certain procedural rules are available to produce conclusive results.