Thirty-three 7th-grade mixed-ability students and 22 11th-grade honors students were interviewed about their conceptions of models and their use in science. Three analyses are presented in order to: (1) portray the character of students' spontaneous answers; (2) examine the criteria students use to decide whether specific items are models or not; and (3) describe how different general levels of understanding models reflect different epistemological viewpoints. Four experts were also interviewed for purposes of comparison. We found that students in both groups have conceptions of models that are basically consistent with a naive realist epistemology. Thus, they are more likely to think of models as physical copies of reality that embody different spatiotemporal perspectives than as constructed representations that may embody different theoretical perspectives. As student ideas become more sophisticated, however, they increasingly include the fact that models are designed for particular purposes, especially to help communication. All of our experts expressed ideas consistent with a constructivist framework, drawing a distinction between abstract and physical models and articulating ways that models are used for the construction and testing of ideas. Our findings suggest that students need more experience using models as intellectual tools, more experience with models that provide contrasting conceptual views of phenomena, and more discussions of the roles of models in the service of scientific inquiry.