Research on "epistemological beliefs" has made important contributions to education, most fundamentally in identifying epistemology as a category of informal knowledge that may play a role in students' knowledge, reasoning, study strategies, and participation. A perspective on students as having epistemological beliefs can provide an alternative interpretive lens for teachers to use in understanding their students' ideas and behavior, in assessing students' abilities and needs, and in adapting their plans and strategies for instruction.

This value of the perspective in opening a new category of instructional perception and intention has not been sensitive to the details of the perspective. At this point, however, further progress will depend on those details. In this chapter, we contend that current perspectives on epistemologies are problematic in their form, or "ontology." We will focus our arguments primarily on epistemological beliefs about science and science learning, specifically in introductory physics, but our contentions apply to beliefs in other disciplines as well.

The study of epistemologies has largely emulated the study of "conceptual understanding," which has established that students have informal knowledge, about physical phenomena for example, that strongly affects what they learn. It is not clear, however, with respect either to epistemological beliefs or to conceptual understanding, how best to model what takes place in an individual's mind. What is the internal form of an informal epistemology? In other words, tolerating a cumbersome alliteration, what ontology should we ascribe to epistemology?

For the most part, researchers have presumed an ontology of "beliefs" as essentially unitary components of essentially stable epistemologies. By "unitary," we mean that each belief corresponds to a unit of cognitive structure, which an individual either does or does not possess. Construed in this way, epistemological beliefs are analogous to the "concepts" or "conceptions" posited as elements of cognitive structure. Just as cognitive science has understood naïve physics to be made up largely of "misconceptions" (e.g. "motion requires force") that differ from expert conceptions ("acceleration is caused by force"), research on epistemologies has understood students to have "misbeliefs" (e.g. "scientific knowledge is certain") that differ from expert beliefs (e.g. "scientific knowledge is tentative"). It follows that, just as developing an understanding of Newtonian physics requires "conceptual change," developing a more sophisticated epistemology requires changing beliefs. In neither case could the naïve constructs — the misconceptions or misbeliefs — be understood to contribute to that development, because they are inherently inconsistent with expert thought.

Therein lies a fundamental theoretical difficulty, as Smith, diSessa, & Roschelle (1993/1994) have argued with respect to the misconceptions perspective: Although couched as "constructivist," it offers no account of productive resources for the construction of more sophisticated understanding. If conceptions are unitary elements of cognitive structure, then student conceptions must be replaced by expert conceptions. To suppose that they could evolve in some way is to suppose, at least, that they have some underlying structure, that they contain productive elements at a finer grain-size; but the misconceptions perspective has not described what that underlying structure might be. Similarly, in considering naïve epistemologies to be made up of constructs such as "knowledge is certain," current perspectives on epistemology offer no account of what may be the raw material from which students could develop new structures, such as that "knowledge is contingent on context and perspective."

There is an empirical difficulty as well. To model students' understanding of physical phenomena in terms of misconceptions is to imply a consistency in their reasoning that is contradicted by evidence from interview protocols (Viennot, 1985; diSessa, 1993; Smith, diSessa, & Roschelle, 1993/1994). Similarly, to presume students' epistemologies exist in the form of beliefs as stable structures is to presume a consistency across contexts: If a belief such as "knowledge is certain" exists as a unitary component of a stable epistemology, an alternative component to an expert's "knowledge is tentative," then this difference between the student and the expert should be consistently evident.

This ontology, which we will refer to as "unitarity," has not to our knowledge been explicitly defended in the literature on epistemologies. It is a default presumption, not the result of a deliberate process of investigation, and we contend it is inadequate. We will argue in the following section that researchers are presuming unitarity, more specifically the consistency it implies, in studying students' beliefs through questionnaire and clinical interviews that are far removed from the contexts of learning in which the beliefs are supposed to apply. This assumption of consistency across such diverse contexts is neither evident nor, when made explicit, plausible. We will then sketch a framework of epistemological resources at a finer grain-size than unitary beliefs, analogous to diSessa's (1993) account of phenomenological primitives at a finer grain-size than unitary (mis)conceptions.

By "unitarity" here, it is important to specify, we are not refering to the idea that epistemological thinking develops in unidimensional stages, advanced by Perry (1970), King and Kitchener (1994), and others. Schommer (1990), Hofer and Pintrich (1997) and others have challenged this assumption, arguing that epistemologies are better modeled along multiple dimensions. We concur. However, as we now elaborate, even authors who reject stage theories of epistemological development continue to assume unitarity in the sense we are using the term, the idea that personal epistemologies take the form of theories or traits.