Thought Experiments

Thought experiments are used as catalyst for boosting ideas and development of theory. Outcome of thought experiment doesn't have to be unambiguous and usually can be paradoxical or unintuitive. The only purpose of such activity is to raise questions, lay some "emotional foundations" for discussed theory at hand. And Btw, thought experiments are used in many disciplines, not only in Physics, they are used in Philosophy, biology and others. Some famous examples, Schrödinger's cat - illustrates quantum laws weirdness and non-applicability to macroscopic world objects; Chinese room,- discusses inability to algorithmically construct and verify strong AI. Some thought experiments may be verified, some - not due to present state of technology, and some serves just as illustration of paradox.

The purpose of a thought experiment is usually not to come to a conclusion, but to clarify thinking on an issue. A thought experiment is an extremely simplified imagined situation--far simpler than could be achieved in real life [1]. A simple situation is easy to reason about because we can have complete knowledge. In this way, it's more like math than science.

A classic thought experiment comes out of the following question: Do heavier things fall faster than lighter things?

Galileo proposed a thought experiment. Suppose you have a small coin and a large anvil. It is natural for one to assume that the anvil, when dropped, will fall downwards faster than the small coin. The anvil's heaviness attests to its "desire" to reach the ground (as thought by Aristotle), which is of much greater magnitude than the coin.

Now consider what will happen if the coin is welded to the anvil and the result is dropped. The coin does not fall as fast, so it must slow down the fall of the anvil, much as a fast horse that is yoked to a slow horse cannot run as fast. So, the acceleration of the welded whole must fall slower than the anvil alone. But, doesn't the welding create a single, heavier object? The increased heaviness must then result in even faster falling.

The contradictions in this thought experiment can only be resolved if all objects fall at the same rate when dropped (in a vacuum, see [1]).

Einstein was particularly fond of thought experiments and often used them to guide his thinking (or to annoy quantum physicists).

Other thought experiments reveal contradictions in current theories. One that is a particular source of arguments among physicists is called the Firewall. Nobody knows how to resolve this one yet.

[1] This is where the joke about physicists studying spherical cows comes from. A sphere is a much simpler shape than the actual shape of a cow.

You do a thought experiment every time you solve a textbook problem. Thought experiments use existing theory plus logic to make predictions. If the existing theory is right, then the results of the thought experiment and of the (eventual) real experiment must match up. Usually the theory is right. This makes thought experiments useful first because they make predictions that are probably right and second because they occasionally make predictions that are wrong, which alerts us that the theory needs patching up.

Feynman appears to have written this part of his textbook without knowing that the experiment had actually been performed (see my comment on the original post). He did the thought experiment, and was extremely confident that the prediction was correct because he was extremely confident that the theory was correct. His confidence turns out to have been justified.