The Next Generation Science Standards (NGSS) emphasize engaging students in science practices through which students can develop content, procedural, and epistemic knowledge of science. To accomplish this goal, science teachers must embed science content in authentic science practice. Successfully integrating content and practice will require significant epistemological and pedagogical shifts on the part of teachers. However, teachers cannot meaningfully achieve this integration if they, themselves, do not understand the epistemic underpinnings of the science practices. Given the lack of literature specifically addressing teachers' epistemic understanding of science practices, we conducted a qualitative study to fill this gap in the literature. Data were collected using an open-ended online survey, and responses from 128 science teachers were analyzed using the constant comparative method. Our results indicate that relatively few teachers exhibited more nuanced epistemic understandings of scientific practice that extend beyond the rigid, linear scientific method presented in science textbooks. Teachers broadly believed that scientific inquiry must begin with a question, seldom cite methodological differences when comparing two experiments, frequently viewed models as teaching tools for representations and explanation, and hardly discussed computational thinking in terms beyond data analysis. More importantly, teachers viewed that scientific conclusions are heavily influenced by researchers' perceptions and biases. To enhance science teachers' sophisticated epistemic understanding of science, we highlight the need not only to engage teachers in authentic scientific inquiry experiences, but also to provide opportunities to learn how to effectively integrate science practices as a coherent system, especially computational thinking and models, into science lessons.