Introduction: This paper explains the development and testing of a novel paper-and-pencil measure designed to support inferences of five epistemological belief dimensions based on three forms of evidence. The three evidence types are anticipated to converge at a single theoretical level, permitting better-supported inferences than existing survey measures. Method: Data were collected from 12 university-level students from Hong Kong (19 to 34 years of age, M = 24.9, SD = 4.2). Participants completed a survey designed to capture three different sources of evidence of epistemological beliefs: epistemological judgments, intentions, and explicit beliefs. A theoretical level (i.e., dualism, relativism discovered, and contextual relativism) is assigned for each evidence type for each of five dimensions. The theoretical-level assignments across the three evidence sources are then analyzed for convergence. Results: Results indicate the three assigned theoretical levels for each dimension did not converge. Therefore, a comprehensive theoretical level per dimension could not be given. Instead, summative scores could to be determined based on average performance per dimension and for all dimensions combined. As a group, participants scored moderately on all dimensions combined, performing worst on the "Organization of knowledge" and "Justification of knowledge" dimensions and best on the "Stability of knowledge" dimension. Discussion and Conclusions: Three explanations for evidence type variability are discussed: a) the MESSEB is incorrectly eliciting the intended evidence, (b) the types of evidence are not appropriate for inferring the intended epistemological beliefs, and (c) epistemological beliefs are being measured as intended, but variation is genuine and typical. The ramifications of each explanation are discussed along with possibilities to address them in future research.