This study describes the possible variations of thought experiments in terms of their nature, purpose, and reasoning resources adopted during the solution of conceptual physics problems. A phenomenographic research approach was adopted for this study. Three groups of participants with varying levels of physics knowledge--low, medium, and high level--were selected in order to capture potential variations. Five participants were selected within each level group and the study was conducted with fifteen participants in total. Think aloud and retrospective questioning strategies were used throughout the individually conducted problem solving sessions to capture variations in the participants' thinking processes. The analysis of the data showed that thought experiments were actively used cognitive tools by participants from all there levels while working on the problems. Four different thought experiment structures were observed and categorized as limiting case, extreme case, simple case, and familiar case. It was also observed that participants conducted thought experiments for different purposes such as prediction, proof, and explanation. The reasoning resources behind the thought experiment processes were classified in terms of observed facts, intuitive principles, and scientific concepts. The results of the analysis suggested that thought experiments used as a creative reasoning tool for scientists can also be a productive tool for students. It was argued that instructional practices enriched with thought experiments and related practices not only reveal hidden elements of students' reasoning but also provide students opportunities to advance their inquiry skills through thought experimentation processes.