_F_ alone of the manuscripts collated offers the correct _audisse_ (for _audire_) at x 17. At xi 21 it and _B_ alone have the correct _mihi_ for _tibi_ (omitted by _C_). At xiv 7 it has the probably correct _muter_ for _mittar_, also found in _Bodleianus Canon. lat. 1_ and _Barberinus lat. 26_, both of the thirteenth century. With the exception of _muter_, these readings could have been recovered by conjecture; given the separative interpolation at iii 44, _F_ differs surprisingly little from the other manuscripts. _H_ The thirteenth-century _Holkhamicus 322_, now _British Library add. 49368_, contains (with _I_) the correct _hanc_ at i 16, the other manuscripts having _ha_, _ah_ (_B_), or _a_ (_C_). At xvi 30, where I have printed _leuis_, the reading of most manuscripts, _H_ has _leui_, the conjecture of Heinsius; Professor R. J. Tarrant informs me that the same reading is found in _Othob. lat. 1469_. At iv 45 _H_'s _qua libet_ is the manuscript reading closest to Heinsius' correct _quamlibet_; most manuscripts have _quod licet_. Most other variants in _H_ are trivial errors, although there seems to have been deliberate scribal alteration at x 18 (_sucus amarus erat_ for _lotos amara fuit_), xiv 38 (_Celsius_ for the usual _Septius_; Scaliger restored _Scepsius_), xvi 3 (_ueniet_ for _uenit et_; presumably the intermediate step was _uenit_), and perhaps at xiv 31 (_miserabilis_ for _uitabilis_). _I_ The thirteenth-century _Laurentianus 36 32_, Lenz's and André's _m_, has the correct _perstas_ at x 83 for _praestas_; its reading is also found in _P_ and as a variant of _F2_. At vii 1 it shares with _C_ the reading _horas_ (=_oras_), which I have printed in preference to the usual _undas_. At viii 15 _I_ has the hypercorrect _nil_ for _nihil_, and at xiii 26 _ethereos ... deos_ for _aetherias ... domos_, but in general has few signs of deliberate alteration. _L_ _Lipsiensis bibl. ciu. Rep. I 2° 7_, of the thirteenth century, has _haec_ at ix 103 for the other manuscripts' _et_. _Haec_ restores sense to the passage, and was the preferred reading of Heinsius; I consider it a scribal conjecture, now rendered obsolete by Professor R. J. Tarrant's more elegant _quae_. _L_'s text has clearly been tampered with at xiv 41 (_populum ... uertit in iram_ for _populi ... concitat iram_), but in general seems to have suffered little from interpolation. It is, however, of little independent value as a witness to the text. _T_ _Turonensis 879_, written around the year 1200, was first fully collated by André for his edition; Lenz had earlier reported its readings for IV xvi and part of I i. At ix 119 only _T_ and _F2_ of the manuscripts collated have the correct _laeuus_, although other manuscripts come close, and the reading could have been recovered by conjecture. At xv 40 _T_ reads _transierit saeuos_ for _transit nostra feros_; clearly _nostra_ was at some point lost from the text, and metre forcibly restored. _P_