Research on personal epistemologies (Hofer & Pintrich, 2002) has mostly conceptualized them as stable beliefs or stages of development. On these views, researchers characterize individual students' epistemologies with single, coherent descriptions. Evidence of variability in student epistemologies, however, suggests the need for more complex models. Hammer and Elby (2002) proposed modeling personal epistemologies as comprised of manifold epistemological resources. This difference in ontology--the form research attributes to cognitive structure--accounts for variability: The activation of these epistemological resources depends on context. Our purpose in this article is to argue that it also accounts for coherences in student epistemologies, in particular for multiple local coherences. We advance this argument using a case study of a 15-min discussion by a group of eighth graders about the "rock cycle" (the cyclic transformations of rock among different forms). We begin with evidence of the students' working from a stable, coherent epistemological stance. Then, after a brief, purely epistemological intervention by the teacher, the evidence indicates they are working from a different but also coherent and stable epistemological stance.