The _fragmentum Guelferbytanum_, _Cod. Guelf. 13.11 Aug. 4°_, generally dated to the fifth or sixth century, is the oldest manuscript witness to any of Ovid's poems. Part of the collection of the Herzog August Bibliothek in Wolfenbüttel, it was discovered by Carl Schoenemann, who published his discovery in 1829; details of his monograph will be found in the bibliography. The two pieces of parchment are a palimpsest, having been reused in the eighth century for a text of Augustine; later they were incorporated into a bookbinding. As a result of this treatment they are in extremely poor condition. _G_ contains all or part of ix 101-8, ix 127-33, xii 15-19, and xii 41-44. To make it perfectly clear when _G_ is a witness to the text, I have not grouped it with other manuscripts, but have always specified it by name. If _G_ is not mentioned in an apparatus entry, it is not extant for the text concerned. _G_ is written in uncial script, with no division between words but with indentation of the pentameters. Its one contribution to the establishment of the text is at ix 103, where it reads _quamquam ... sit_ instead of the more usual _quamquam ... est_ found in the other manuscripts. In general, the text offered by _G_ is surprisingly poor. At ix 108 it reads _fato_ for _facto_, at ix 130 it has the false and unmetrical spelling _praeces_, at ix 132 it has _misscelite_ for _misi caelite_, at xii 17 it reads _lati_ for _dilati_, and at xii 19 _naia_ for _nota_. These errors demonstrate that the rest of the tradition does not descend from _G_. Korn gives an accurate transcription of the fragment in the introduction to his edition; photographs of parts of the fragment can be found at Chatelain, _Paléographie des classiques latins_, tab. xcix, 2 and E. A. Lowe, _Codices Latini Antiquiores_, vol. IX, p. 40, no. 1377. _B_ and _C_ _Monacensis latinus 384_ and _Mon. lat. 19476_, both dated by editors to the twelfth century, are descended from a common ancestor. This is easily demonstrated by the large number of shared errors not found in other manuscripts[9]. At iv 36 _B_ and _C_ have _intendunt_ for the correct _intendent_, at viii 6 _uolo_ for _uoco_, at viii 18 _perueniemus_ for _inueniemur_ (_-ntur_,_-mus_), at viii 44 _illa_ for _ulla_, at viii 89 _cara_ for _care_, at ix 44 _fingit_ for _finget_, at ix 71 _quod_ for _cum_ (_FILT_) and _ut_ (_HM_), at ix 92 _praestat_ for _perstat_, at ix 97 _et_ for _ut_, at xiii 5 _certe est_ for _certe_, and at xiv 30 _culpatus_ for _culpatis_. In some of these passages _B_'s still visible original reading has been corrected by a later hand. In other passages it is clear from the signs of correction that _B_ originally agreed with _C_ in distinctive readings now preserved in C alone: _subito_ for _sed et_ (iii 27), _erat_ for _eras_ (vi 9), _occidit_ for _occidis_ (vi 11), _suspicit_ for _suscipit_ (ix 90), _parent_ for _darent_ (xvi 31).