As a consequence of the COVID-19 pandemic, most courses at a large technical university were adapted so that students had a free choice of whether to attend lectures on-site or online; in addition, in many courses, lecture recordings were available. At the subsequent exam session, over 17,000 student-survey responses were collected regarding attendance choices, learning behavior, interest in the course, perception of the exam, and recommendations to future students. A total of 27 learner attributes and their relationships were investigated. In addition, conditional attributes and free-response statements were analyzed, and the students' exam grades were retrieved to gauge their performance. We found only minute differences with respect to exam performance, but the analysis indicates distinctly different preferences and constraints in taking advantage of learning opportunities. We also found some indications that performance differences might be larger for interactive-engagement courses. The results of the analysis may be key to answering why at many universities, faculty report that live-lecture attendance has decreased more strongly than expected with the availability of new, virtual attendance modes.