The purpose of this paper is to determine how the following manuscripts of Cicero’s De oratore relate to each other: Avranches 238 (A), Erlangen 380 (E), Leiden, Voss. Lat. O. 26 (F) and Vatican, Reg. lat. 1762 (K; a florilegium). In the first section it is shown, on the basis of conjunctive and separative errors, that E and F must have had a common exemplar (α). Next, I argue that E and F (and hence also α) are derived from A (Section 2). In Section 3 I turn to K and argue that K’s excerpts must have been taken from two different exemplars: §§ 429-41 from α, §§ 401-28 from some related manuscript, probably A itself. Section 4 is devoted to a few questions of secondary importance. Finally, I give a stemma.