$86. (Rutilii ?) rutili BV.

§ 88. (immolavisse ?) immolasse PV.

§ 91. (judico) judicio PV1.

§ 93. (gentis) sentis PV.

(contemnet) contempnet PV.

This gives 13 agreements between V and B, 12 between V and

P, 7 between V and C.

(2) Agreements between A and any other single Ms.

§ 47. (omniaque quae) omnia quaeque AC.

(accipitres) accipitros AP and perhaps V'.

§ 49. (Erechtheus) erectheus AB.

This gives 8 agreements between A and B, 12 between A and C, 3 between A and P as contrasted with 55 agreements between A and V.

I proceed now to examine the readings in which the accepted text rests on the authority of a single codex.

True text preserved by V alone.

§ 11. credis esse, where A has credidisese, B credissesse corr. in credidisses, CP credidisse, credisse E. Here the variety of readings suggests a fault in the archetype: if it had credidsesse, this might easily be corrupted into the other readings and corrected in V.

$ 20. a consuetudine enim, where enim is omitted by the other MSS and given in contraction by V. In this case there can be little doubt that V represents the archetype.

$ 35. quem ipsum non omnes interpretantur uno modo, qui quoniam quid diceret intellegi noluit, omittamus, where the other MSS. omit qui and even V seems uncertain. Baiter thinks non omnes-modo to be a gloss, which would naturally suggest a connecting qui before quoniam on the other hand qui would easily be lost before quoniam.

§ 62. talis a philosophia pellatur error, where the final a of philosophia (which probaby commenced a new line in the archetype) is written separately in V'. From this the readings of the other MSS would easily spring, viz. t. a philosophiappellatur error A, t. a philosophi 'appellatur e. B'C, t. philosophia appellatur e. B2 (evidently altered to make sense), and so t. a philosophis appellatur e. PE and V2 (only that the last has apellatur).

§ 70. quisquamne istuc negat by corr. in V, where A has quisquam iuste (iuste being 'in ras.' by a late hand), C quisquam istuc, BP quisquas istuc (quas cancelled in B), quid istud E. I am inclined to think the archetype had quisqua stuc (for exx. of stuc in Cicero's MSS see my n. and Lachmann's Lucr. p. 197), so that here the unaltered B and P are its best representatives.

§ 73. neque ut inde auferam (see n. in loc.) which, written. continuously in the archetype, is nearly the same as V' neque tinde auferam; the other readings are easily explicable from V, viz. neque unde auf. ACEP, and ne quid inde auf. B, which may be compared with the superscription of quid over neque in V.

§ 84. pecunia edixisse V, where B' has pecuniae dixisse, and AB'CP pecunia dixisse. Here it would seem that B must be either taken directly from a мs in which the words were not separated, or (as I think more probable), from a мs wrongly copied from such an archetype.

To these some would add § 66, where V has permiciem see in loc.

True text preserved by A alone.

§ 67. posquam (so Baiter), others postquam.

§ 68. quem clam Thyestem, see nn. P has quem cleanthyestem, C and by corr. B have quendam thyestem, V quem dant hyestem, (corr. fr. hyestim), E quam dant thiestem. Here there can be no

doubt that A represents the archetype, that in CB and V cl have been mistaken for d and that in V the letters were wrongly grouped. This seems to show that A is independent of V.

§ 11. Sagram A, all others are allured by the easier sacram.

§ 22. inanimarum AB2V2, inanimatarum B'V'E, animarum C. There can be little doubt that the archetype (a) is represented by A and that the commoner form is written by error in B'V1.

§ 48. duces A', right as shown by the following respondebis, B'CEV' have by carelessness ducis, PA'V' dicis, B2 dices.

§ 71. commemorabantur A, commemorabatur others.

§ 89. quidam A (so Baiter), quidam amicus the other MSS, except that B has quidem. It seems more likely that A should have overlooked amicus than that it should have been inserted by the rest.

True text preserved by B alone.

§ 4. parum, parvam AV1E, parva CV. Probably a had parū, which being followed by accepi would suggest parva or parvam: or, if Madvig is right in reading cepi, accepi may have arisen from parua cepi in a, in which case B's reading would be an emendation.

§ 8. quod esset perspicuum BV3, where AV1 have q. est p., and CE q. et p. probably by mere carelessness.

$11.

eos tu cantheriis: here V has tuq., P tuque, AC tu quae, E que tu. I am inclined to think that the q. of V (and probably of a) was intended to be the first letter of cantheriis, spelt with qu for c, like quotta for Cotta, quoinquinari for coinquinari, quohaerere for cohaerere, quorum canium for Coruncanium. We have a similar instance of a word just begun in § 82 1. Platonem legens noticed

above.

§ 13. rationes requiro BV. Baiter with the other мss omits requiro and it was no doubt a natural word to supply, but A's reading of the following word, recuntur for the secuntur of BV, seems to have arisen from requiro sequuntur, the eye of the scribe passing from the 1st to the 2nd qu. If so we must suppose an intermediate link between a and AV on the one side, as between a and B on the other side, the former link having recuntur, changed by V' into secuntur, the latter preserving the reading of a.

§ 18. omniaque quae a te BV, omnia quae a te APV1 (d erased after a in AV), omnia quaeque a te C, omnia que a te. Here APV

must either represent the archetype, emended by V3 and B, or else the former are derived from the same incorrect copy of a.

§ 24. fieri non possunt B', all others nonne. There can be little doubt that non is what Cicero wrote, but the reading nonne is not so objectionable as to call for emendation. It would seem therefore that B here represents the archetype, and that the others are connected with it by a medium which in this instance proves to be less trustworthy.

§ 26. Orionem B, orationem ACEPV. Here in all probability a had a contraction wrongly interpreted by all but B*.

§ 27. cientis [B], scientis ACEPV. This is another evidence showing that B is connected with the archetype by a different line of descent from that of the others.

§ 42. (Lysithoe est), Lysitho est B (perhaps Lysithoest in a), Lysito est ACPV, lisito est E.

§ 43. (capedunculis iis), cap. his B, om. iis ACEPV. The demonstrative seems necessary, but was of course easily lost after -lis.

§ 47. ibis B, ibi AEV1, ibi C, ibes V3, nothing said of P. Here C appears to represent a, while AV have neglected to mark the

abbreviated s and B has written it out in full.

§ 49. si sunt hi di BE, si sunt di A', si sunt id V1, si sunt ii dii C, si hi sunt di P, si sunt hii di A3, si sunt hi dii V3. Probably a omitted the demonstrative, as in § 43 si di sunt, where I have added isti. If Cicero wrote si hi di sunt, it would be easy for hi to drop out, and the variety of readings suggests that each scribe reinserted it, where he thought best.

$52. jam B, tam CEV, and probably A', tum A', nothing said of P. Here B either follows a separate tradition or has emended the common archetype.

mare B, mater ACEPV. An abbreviated mater might easily be mistaken for mare and v.v. The readings may be most easily explained by supposing two recensions of a, one followed by ACPV (B), the other by B (7).

§ 54. Mnemosyne [BP], nemosine E, nemo sine ACV (nemo 'in ras.' A) †.

§ 57. Cynosuris [BP], gynosuris ACE, ginosuris V'.

§ 66. exitium BE, exitum ACPV. It seems more probable that

* Deiter (Rh. Mus. 1882 p. 314) states that B has orationem like the rest. + Deiter gives mnemosine as B's reading.

the 1st recension (B) should have gone wrong than that B should have corrected exitum.

$75. (abiegnae) abiegne by corr. in B, abigne all others.

§ 82. Anaxarchum [BP] anxarcum A'CE, anxarchum V.

§ 83. praedo felix habebatur B, p. filia h. ACEPV, the inferior MSS have fulia, filica, summus, in Pamphylia &c. I have sometimes thought that felix and filia might both be corruptions of γαζοφύλαξ, that being the office held by Harpalus at the court of Alexander, but if so, several words must have been lost. It does not seem possible that felix should have been an emendation of filia, so that we must in any case recognize here again two recensions of a.

manubiis, BEC, manubiis is AP, manubiis iis V, manibiis C'. Here I should think the reading of AP is that of the 1st recension of a, arising from dittographia of -is, V's reading would be a natural correction of this.

§ 84. quod quisque sacri haberet B (so Ba.), q. q. a sacris h. ACEPV. Here it hardly seems possible for the one reading to have grown out of the other by inadvertence, and as there was more to tempt an intelligent scribe to alter the latter than the former reading, I am inclined to consider sacri an emendation.

in tyrannidis rogum B, in typanidis rogum AEPV, in timpanidis rogum C. If my view of the passage is right (see n. in loc.), several words are omitted owing to homœoteleuton; B has preserved tyrannidis, but altered ut into in with all the other мss.

True Text preserved by C alone.

§ 23. saepe dixti C, dixi ABEPV. It is easy to understand an ignorant or careless scribe writing dixi for dixti, but how are we to account for C? Is it directly copied from a, or is it a correction, not needing very much acuteness, of a wrong traditional reading?

§ 41. in monte Oetaeo C, in monte moetaeo AEPV, in monte metaeo B. I have suggested that the prevalent reading may have originated in in montem oetaeum wrongly divided. If so, C's reading is an emendation.

§ 42. accepimus CEV, accipimus others.

An easy emendation.

§ 52. nihil horum CEV, nihil honorum ABV', n. bonorum P. The abbreviations of honorum, bonorum and horum are easily confused, the only question is how C got the right reading. I think by emendation or by copying from V'.