This study addressed two research questions. (1) Is the multitrait-multimethod confirmatory factor analysis the better statistical tool to detect the presence of five distinct dimensions in a traditional survey tool for epistemic beliefs? (2) What are the epistemic belief and achievement goal orientation profiles of students at different levels of mathematics in the junior college system? The sample consisted of 456 and 443 respondents during the fall and winter quarters, respectively. Participants were enrolled in all levels of mathematics classes from a community college district in the Pacific Northwest. An epistemic belief scale was constructed using items from survey instruments in the personal epistemology literature. Along with epistemic belief dimensions (certainty, structure, authority, innate ability, quick learning), the survey items were also classified as being axiomatic, ontologic, deontologic or procedural in nature. The multitrait-multimethod analysis revealed the presence of five distinct dimensions of belief, and accounted for an additional source of variance related to the wording of the items. Fit indices suggest a strong correspondence between model and data. Students' achievement goal orientations were measured with a survey based on Nicholls' original task, ego and work-avoidant framework. Statistically distinct profiles emerged for the four levels of mathematics surveyed (developmental, terminal/non-calculus math, calculus preparation, and post-calculus). Differences between the math levels were detected for all epistemic belief dimensions except structure of knowledge. The strongest differences were between the developmental level and the calculus preparation or post calculus levels. The effect of math level was statistically significant for all three goal orientations. Task orientation was higher for the calculus prep level than for the terminal math level. Ego orientation was lower for the developmental level compared to both calculus levels, and was lower for the terminal level than for the post calculus level. Work-avoidance was higher for the post calculus level compared to the developmental level. The only statistically significant difference for self-efficacy in test-taking was between the developmental level and the post-calculus level. Lack of variation in self-efficacy across the four levels may suggest a self-selection bias in the sample.