^{page 43 note 1} Knecht, D., CIRIS: authenticité, histoire du texte, édition et commentaire critiques (Bruges, De Tempel 1970)Google Scholar.

^{page 43 note 2} Knecht draws up lists to illustrate these points on pp. xxxiii-xxxvii; his ideas of a true reading are sometimes a little bizarre but the general points stand. H² is also systematically recorded in the apparatus of Vollmer's Teubner edition of the Appendix.

^{page 43 note 3} Cf. the following instances where H² provides an erroneous correction for a wrong reading in H: 106 Alleato H: Allratoe H², 210 erectus H: ereptis H², 249 seonia H: eonia H², 423 dereptis H: deraptis H²; and the following where H² alters a correct reading in H: 22 quale, 93 cupide, 137 magni, 288 nam (for nate), 317 luctu, 521 uersas. Some of these will naturally appear less significant than others. We should note too H^2's superscript comment at 224 uel famulus for H's famulo.

^{page 43 note 4} H² does share many corrections and some errors with x (Knecht's lists pp. xxxiv-xxxvi) but such agreements could, I suppose, be all fortuitous as Knecht believes (p. xxxvii).

Interesting is H^2's change at 210 ereptis for erectus in H (adduced in note 3). For one other manuscript does have a comparable error, and that is L (arreptis), a single manuscript of the x group.

^{page 44 note 1} 17 possim, 109 solitum, 376 talo, 392 thetis, 420 aduxi, 451 pisces.

^{page 44 note 2} See Clausen, W. V., CPh LIX (1964), 90 fGoogle Scholar. for a convenient account of the discovery of HARL (and GB) and some indication of the way editors have viewed them.

^{page 44 note 3} See Knecht pp. lxxix–lxxxi.

^{page 44 note 4} They are described pp. xlviii–lxii, evaluated pp. lxv-lxxxiv.

^{page 44 note 5} I think more work on the inter-relationship of the x manuscripts might show that one or two are drawing on each other; but in view of the fact or probability of contamination (as mentioned above) such work would be much obscured and still could not totally eliminate any of them from consideration. Cf. below p. 47.

^{page 44 note 6} As, I think, a look through Knecht's apparatus quickly shows. For example see how L and Ch. at 370 have picked up R's interpolation contunditi and some of the passages in note 1 on p. 47 may be relevant. On Cors, see below.

^{page 44 note 7} Most editors and commentators seem to have assumed that these interpolated manuscripts (or the one they choose) derive from Z. See the apparatuses of Vollmer and Goodyear for example; also the discussion already cited of Clausen. In case readers think Knecht has given a plain answer, it should be made clear that his use of the word branche in this introduction (where there is no stemma) is bewilderingly vague and varying.

^{page 45 note 1} See Housman, , CR XVII (1903), 304 Google Scholar – though Housman refers only to L.

^{page 45 note 2} Z (wrongly, pace Knecht) reads laudate. The correct gaudete was restored by Schwabe. The most obvious explanation of both corruptions seems to me to be as follows: the original gaudete was simply omitted by haplography. The scribes of both Z and x were faced with a lacuna; each filled it in as seemed best to him – the scribe of Z perhaps echoing the Psalmist, as Alton in Hermathena XLII (1920), 79 Google Scholar suggests, the scribe of x padding with a second epithet. It seems very unlikely to me that x would have interpolated blandaeque if faced with laudate: other wild interpolations seem to have more provocation – thus supposing x was faced with uenus illi at 471, the nonsense or apparent nonsense of it is some explanation for the scribe's interpolation strop(h)adasque; the same applies to numina seruant of x at 318 for fatare seruant, the probable reading of his exemplar.

^{page 45 note 3} summus Z. I am aware that it is quite possible that x simply interpolated hinc sinus from 471 for a meaningless summus. But perhaps a more likely explanation of the variants is that behind Z and ξ (as we may call the ancestor of x) lies something like sinius which split into two separate corruptions sum(m)us and sinus; then hinc would very soon be added to the latter to complete the metre. With this account may be contrasted what Knecht has to say on p. lxxxiii.

^{page 45 note 4} Not accepted by Knecht as the true text.

^{page 45 note 5} Considered a probable example of preservation by Knecht.

^{page 45 note 6} It would of course be a palaeographically simple correction from emicleos (amicleos AR) pergens Z; but, since the sense of the line is so difficult, certainly not compelled by context.

^{page 45 note 7} We note that there would be no help for an emendator in 204; in any case the reading of x here scarcely sounds like an emendation.

^{page 46 note 1} Many editors (including Knecht) adopt munera; I think this impossible but will save my explanation for the commentary I am preparing on the poem.

^{page 46 note 2} 249 seonia H: om. A: morbo R; 264 mediis ex hostibus HA: mediisque ex hostibus R; 326 seua HA: scaeua R; 370 incendit HA: contundit R; 413 e terris HA: terris R (with G) – terris is probably nearer the truth than e terris, but I suspect that Z read e terris and R omitted e to restore metre.

^{page 46 note 3} Here Knecht treads with admirable wariness. Since x is prone to interpolation not all variants between H and AR will be suitable evidence for determining the affinities of x. Knecht includes only cases (a) where x parallels an error of H against the truth in AR or vice versa; (b) where the variants of H and AR have more or less the same value (either being equally plausible or equally nonsensical) – i.e. cases where a scribe faced by one of the readings is unlikely to have wished or been able to interpolate the other.

^{page 46 note 4} Knecht p. lxxii; Clausen op. cit. p. 93.

^{page 46 note 5} That sidera (Hx) was the paradosis, indeed the truth, at 50 needs some supporting, but I reserve my discussion for my commentary. I pause also over 226, but I have not yet satisfied myself exactly what either the truth or the paradosis of this line was.

^{page 47 note 1} Cf. p. 44 note 6 above. We might consider at this point the passages adduced by Clausen in support of the following proposition (op. cit. p. 93): ‘Since L [the only one of the x manuscripts he considers] agrees with H in a few wrong readings that can hardly be owing to coincidence or conjecture, it may also be derived from H [as he calls the exemplar of H].’ The passages he refers to are 17, 45, 117, 199, 226, 273, 334. In 17 and 273 the errors of H (possum and te per) are in fact shared not just by L but by x; possum in 171 should think was in fact quite probably in the exemplar of Z and ξ – possim of AR will then be either a mechanical error or an attempt at emendation; in 273 te per of Hx admittedly looks like a normalization, but I do not think we can say with much certainty what the paradosis was. 45 exegimus ANUCh.: exigimus HCors.L (the line is omitted by R): it does look as if Cors, and L have been infected with H's error. 117 infestos ad muros HUCh.: infesto ad muros ARN: infestos muros L: perhaps the likeliest explanation of the corruptions would be that Z and ξ read infestos ad muros, ARN managed to make the obvious correct change, while L merely restored metre. 199 uox o H: uos o AR: uox Cors.: uix o NLU: uis o Ch.: probably all the readings of the x manuscripts derive from uoxo; perhaps thus they are dependent on an error of H. But the disparity in their readings might suggest the effect of a long period of transmission; it might suggest in other words that ξ and Z both read uox o and that AR's correct uos o was a result of simple emendation. About 2261 reserve judgement (see previous note). In 334 (non ARNUCors.: nisi HLCh.) it does seem that L and Ch. have got their reading from H.

^{page 47 note 2} His deductions are based on lists of occasions (a) when x varies from Z with the exception of one or two manuscripts, (b) when x agrees with Z with the exception of one or two manuscripts, (c) when x splits between variants in H and AR.

^{page 47 note 3} Cf. above p. 44 n. 5.

^{page 48 note 1} The basic ericth- of the x group at 66 seems an attempt to emend some sort of sense into a transmitted erith-. That Cors, has a final s could conceivably mean that it goes back to a reading eritheis i.e. something closer to crateis, but we can hardly build upon it.

^{page 48 note 2} montis abisses is possibly the truth (suggested by Scaliger); marginally perhaps – other things being equal – montis iisses is more attractive. This is the sort of case where manuscript authority could be important: if Cors, can be shown to have independent worth, I should be inclined to adopt montis abisses. The unmetricality of Cors.'s reading could, as Knecht suggests, be significant.

^{page 48 note 3} See above, p. 45.

^{page 48 note 4} Knecht is unclear on the point whether the correction is by the first or a subsequent hand: in his app. crit. he writes ‘hinc sumus (i. sup. m add. m. ut uid. scholiastae) Cors.’; on p. lxviii he talks of ‘sumus corrigé en sumius Cors.’. I doubt actually that a decision could be reached without an autoptic examination of the manuscript; but for the conclusion reached above, essentially concerning the manuscript's value, it does not much matter.

^{page 48 note 5} Both Vollmer and Knecht record a change of mind in R from thetis to thetys. One is tempted to see this in terms of possible access to purer tradition – cf. above, p. 46. But here the benefit of Knecht's minute account of the manuscripts can be seen: substitution of y for i is a constant habit of R (pp. xlvii–xlviii) – as it was with numerous other scribes.

^{page 48 note 6} Thus it might be that in 66 L is emending on the basis of Cors.'s irrational, and hence possibly ‘sincerar’, reading.

^{page 48 note 7} Cors, does not (it may be worth reminding ourselves) differ significantly from any of the distinguishing x readings I mentioned above on p. 45, except at 472.

^{page 48 note 8} We remember the AR readings it carries (above, p. 47); and cf. also how Cors, carries what is a peculiar variant of H at 45 : cf. note 1, p. 47.

^{page 48 note 9} I seem thus to see Cors, a little differently to Knecht: ‘Le copiste du codex Corsinianus semble donc avoir puisé à deux sources différentes, l'une appartenant à la branche AR de la tradition, l'autre très proche des manuscrits L et Ch.’ (p. lxxxiii).