[larksimian]

Mathematics is based on logical deduction. You start with a set of rules and you prove additional rules.

Science is based on (lots of things but primarily) empirical induction. You observe some phenomena, decide that they are of the same type, generate a model that predicts how all phenomena of that type behave. By necessity you observe an infinitesimal fraction of the phenomena in your type.

In math proofs are universal and timeless and complete. There is no additional observation to be made. A proof can be wrong, but if we conclude that it is logically correct then that's it. (edit: well, there's the incompleteness theorem: it could be that a set of axioms is inconsistent, but then EVERYTHING based on those axioms becomes invalid. It's still quite different than scientific progress which changes models a little bit at a time.)

Scientific models are approximations of reality(very good approximations often!) but they get tweaked as new observations come in. Or just wholesale thrown out(or at least become obsolete), see Newtonian orbital mechanics vs relativistic orbital mechanics. Newton was right-ish, but relativistic solutions are right-er(possibly fully correct?)[1].

With deductive reasoning you are expanding a definition. With inductive reasoning you are guessing at connections between observed events.

[1] http://physics.stackexchange.com/questions/26408/what-did-ge...

[cygx]

Or just wholesale thrown out(or at least become obsolete), see Newtonian orbital mechanics vs relativistic orbital mechanics.

Newtonian orbital mechanics is not obsolete: Within the limits of its applicability, it's a perfectly fine framework and in use to this day.