Current science assessments typically present a series of isolated fact-based questions, poorly representing the complexity of how real-world science is constructed. The National Research Council asserts that this needs to change to reflect a more authentic model of science practice. We strongly concur and suggest that good science assessments need to consist of several key factors: integration of science content with scientific inquiry, contextualization of questions, efficiency of grading and statistical validity and reliability. Through our Situated Assessment using Virtual Environments for Science Content and inquiry (SAVE Science) research project, we have developed an immersive virtual environment to assess middle school children’s understanding of science content and processes that they have been taught through typical classroom instruction. In the virtual environment, participants complete a problem-based assessment by exploring a game world, interacting with computer-based characters and objects, collecting and analyzing possible clues to the assessment problem. Students can solve the problems situated in the virtual environment in multiple ways; many of these are equally correct while others uncover misconceptions regarding inference-making. In this paper, we discuss stage one in the design and assessment of our project, focusing on our design strategies for integrating content and inquiry assessment and on early implementation results. We conclude that immersive virtual environments do offer the potential for creating effective science assessments based on our framework and that we need to consider engagement as part of the framework.