A new approach to analyze scientific methods as patternsof state transitions is proposed and exemplified by the two mostimportant, general methods: induction and deduction. Though only`local' states of science are considered in this paper, includinghypotheses, data, approximation and degree of fit, the approach caneasily be extended to more comprehensive kinds of states. Two `pure'forms of induction are distinguished, enumerative and hypothesisconstruction induction. A combination of these two forms is proposedto yield a more adequate picture of induction. While the pure forms ofinduction are clearly distinct from the deductive pattern, the patternof the combined form of induction is very similar to the latter. Thepresent account of scientific methods not only points out thedifferences between different methods but – in contrast to usualdiscussions of methodology – also clarifies what they have in common.