A mathematician experiments, amasses information, makes a conjecture, finds out that it does not work, gets confused and then tries to recover. A good mathematician eventually does so — and proves a theorem.

Conformal Mappings, American Scientist, Sept.-Oct. 1999 (p. 445)