| Thought experiments are a means of imaginative reasoning with an employment record longer than two and a half thousand years. Used by Aristotle, Galileo, Newton, Darwin, Maxwell, and Einstein, they form part of the education of every scientist alive today.;While most scientific instruments aim to increase the precision of our interaction with empirical data, thought experiments leave the empirical realm behind. They spin buckets in an empty universe, summon demons to play with particles, and challenge us to throw spears at the edge of space. If tools of the imagination like thought experiments are important in science, what role are they playing?;Using the methods of history, philosophy and cognitive science, I argue that while thought experiments do not always discover or justify new facts, they do usually increase the empirical content of theoretical structures (laws, models, concepts, etc.) for an agent. Empirical content is increased when the agent connects a theoretical structure to existing concepts, experiences, values and abilities. Through these connections, thought experiments increase scientific understanding. Knowledge can be produced when understanding is applied, but this is a separate achievement. I argue that the understanding produced by thought experiments is fallible but necessary for scientific progress, in the same way that intuitions about what a speaker means are fallible but necessary for linguistic interpretation.;Chapter 1 introduces the philosophical literature on thought experiments. Chapters 2-4 reject two important accounts of thought experiments. Chapter 5 examines historical cases to focus on the role of thought experiments in science. Chapter 6 considers results in cognitive and social science. Chapter 7 presents a transcendental argument that grounds our ability to increase empirical content through thought experiments in human imagination, and explores some implications of this argument for the relationships among imagination, understanding, explanation, knowledge, representation and objectivity. |
