Recent work in artificial intelligence, or AI, has produced programs capable of serious intellectual work in science. Results from AI are used to show that there exist mechanized procedures for discovering hypotheses and that these methods often lead to plausible hypotheses. The question to be answered is: By what methods can a computer program find plausible explanations of empirical data? To answer this, the common metaphor of search for describing several AI programs is examined followed by a look at examples (...) of programs actually constructed. Five types of search considered are: random, exhaustive, selective, heuristic, and opportunistic search. From there, the paper generalizes to some conclusions about mechanizing hypothesis formation. (shrink)