This article discusses theoretical assumptions either explicitly stated or implied in research on teachers' beliefs. Such research often assumes teachers can easily articulate their beliefs and that there is a one-to-one correspondence between what teachers state and what researchers think those statements mean. Research conducted under this paradigm often reports inconsistencies between teachers' beliefs and their actions. This article describes an alternative framework for conceptualizing teachers' beliefs that views teachers as inherently sensible rather than inconsistent beings. Instead of viewing teachers' beliefs as inconsistent, teachers' abilities to articulate their beliefs as well as researchers' interpretations of those beliefs are seen as problematic. Implications of such a view for research on teacher beliefs as well as for the practice of mathematics teacher education are discussed.