A method is presented for analogical reasoning in Automated Deduction. We focus on the abductive aspects of analogy and give a unified treatment for theorems and non-theorems. Abduction allows to deal with partial analogies thus strongly increasing the application field of the method. It also allows to detect "bad analogies" in several cases. Explanatory examples as well as more realistic examples quantifying the effects of using analogy (for theorem-proving and for counter-example building) are given.