In this article, I present an analysis of the epistemic value of the students' conceptions, as employed in the current constructivist research. I focus on the conceptions about natural kinds. Since natural kind terms are a crucial part of the discourse of the natural sciences, my conclusions are particularly relevant in science education. To perform my analysis, I use a thought experiment, adapted from Hilary Putnam's famous Twin-Earth examples. I conclude that, to avoid some strong ontoepistemic implications, an externalist view of the learner's thought must be adopted. In my approach, students' conceptions are re-interpreted in terms of Hilary Putnam's stereotypes, conventional representations of a natural kind that guarantee a semantic competence in a specific linguistic environment. As a conclusion, I argue that the role of students' conceptions is not epistemic but pragmatic. They should not be understood as a partial or provisional representation of how a natural kind really is but as linguistic tools that permit the learner to engage in classroom discourse. Finally, I present a preliminary application of my ideas to the study of the usage of tautologies in science education.