The process of explaining a piece of evidence by constructing a set of assumptions that are a good explanation for that evidence is ubiquitous in real-life (e.g. in legal systems). In this paper, we introduce, discuss, and formalise the notion of stable explanations in a non-monotonic setting. We show how, while applying it to the process of (1) computing a set of literals able to (2) derive a conclusion (3) from a set of defeasible rules, we obtain a restricted version of the notion of abduction. This is both interesting and useful: when an explanation for a given conclusion is stable, it can, in fact, be used to infer the same conclusion independently of other pieces of evidence that are found afterwards.