On the basis of the results of two prior studies at the US Naval Academy (USNA), which described the choice of study resources and the self-reported learning approaches of students of differing achievement levels, the current investigation examines how students of differing achievement levels in general chemistry actually solve multiple-choice questions. A think-aloud protocol was selected as the vehicle for this investigation. This research utilized and compared the correlation of both a holistic qualitative and a quasi-quantitative approach to analyzing the interviews. The holistic qualitative approach identified student behaviors in four broad categories: problem-solving, conceptual understanding, test-taking strategies, and use of scientific language. The quasi-quantitative analysis allowed us to focus on more specific behavioral trends within these categories providing a more detailed picture of what middle-achieving students do when solving algorithmic and conceptual problems. Middle-achieving students demonstrated more variability when solving conceptual questions as compared to algorithmic questions, applying a mixture of behaviors that were characteristic of higher-achieving and lower-achieving students. Implications for teaching based on this research include the need to help middle-achieving students become aware of the difference between their approaches to solving algorithmic versus conceptual questions, emphasizing what they do correctly and how they can improve their problem solving.