Karl Duncker made a series of empirical studies on problem-solving in 1925-40. The problems were mostly mathematical problems, but many of the findings apply to other kinds of problems as well. For a usability researcher, the most interesting findings relate to thinking aloud protocol that he used in his studies and the transfer mechanisms of problem solving.
Karl Duncker makes a clear distinction between thinking aloud and introspection in his article: as subjects think aloud, they just allow their activity to become verbal instead of explaining the reasons for their actions.
Duncker discusses also on the thinking aloud protocol in his article and claims it to be a reliable source of information on those things that it contains, but reminds that the fact that something is not verbalized, does not mean that it was not in the subject’s mind. As an example, he mentions the phenomenon in which the mediating phases preceding the final solution easily blend into the solution and are left unverbalized.