Aristotle's logical theory is centrally concerned with deductions (συλλογισμοί). A deduction, for Aristotle, is 'an argument in which, certain things being assumed, something else than what has been assumed results of necessity through what has been assumed'. This definition is from the opening chapter of the Topics. Similar definitions are given at the beginning of the Prior Analytics, Sophistici Elenchi, and Rhetoric. In none of these passages, however, does Aristotle explain in any detail what the definition and its individual parts mean. Instead, his most extensive discussion of the definition of deduction is to be found, perhaps unexpectedly, in chapter 6 of the Sophistici Elenchi. This chapter has received relatively little attention in the recent scholarly literature. Nevertheless, it has important implications concerning the nature of deductions-or so I will argue. My aim here is to explore what we can learn from the chapter about Aristotle's conception of deduction. The Sophistici Elenchi deals with apparent refutations, that is, with arguments which appear to be refutations but are not refutations. In chapters 4 and 5 of the treatise, Aristotle identifies thirteen kinds of apparent refutations. In chapter 6 he states that these thirteen kinds can ultimately be reduced to one of them, namely to ignoratio elenchi (see Section 1 below). In order to prove this, he argues that all apparent refutations violate some condition laid down in the definition of refutation. Since refutations are a kind of deduction, his argument also appeals to the definition of deduction (Section 2). Aristotle explains why various apparent refutations violate some condition in this latter definition. In doing so, he appeals to two conditions which are not explicitly included in the standard definition of deduction quoted above. Thus Aristotle extends the standard definition by two new conditions which he does not state elsewhere (Section 3).