^{page 413 note 1} The MS., as we have it, ends at XIV clvii. 1, the last leaf (fol. lvii) having been lost.

^{page 414 note 1} The word gentes appears in contracted form here and at VIII xi. 3, gns (second transcriber) and XII vi. 5 gts (third), each with suprascript stroke to indicate the contraction.

^{page 415 note 1} For example, et is expressed by the first transcriber by a sign like our Arabic numeral 7; while the second uses that ligature that is still preserved in our symbol &c. The third transcriber writes prāre for praestare.

^{page 415 note 2} In f the ‘subscriptions’ cease with Book VII.

^{page 415 note 3} XIII i–iii. were ‘extra ordinem paginarum’ (cf. IX. init.), so that the real beginning of Book XIII, the ‘Xenia,’ is at epigram iv.

^{page 415 note 4} L has preserved for us the fuller form of this clause, viz. ‘in forum (leg. foro) divi Augusti Martis.’ The forum Augusti was styled the forum Martis from the fourth century, according to Jordan, ‘Topography,’ pp. 213, 472; and at the time of this publication the old and the new name were osed in conjunction. In the prefatory Epistle of Book IX the opening was written as part of the ‘subscription’ and is wholly omitted by P, while Q has curtailed it. L has it in full: Ave, mi Torani, frater karissime.

^{page 415 note 5} I cannot understand why Friedlaender should discredit Q's title-headings in these two books. They are as Gennadian as the ‘subscriptions.’ Now that Prof. Maleyn in his recently published work on Martial (in Russian unfortunately) has shewn that all or most of the supposed discrepancies between P and Gruter's Palatinus are non-existent, the ‘consensus’ of P and Pal. goes for nothing.

^{page 415 note 6} A late Latin use of circa, which has received some attention from scholars lately (see Archiv f. lat. Lexicographie, viii, 179; ix, 559) is found in the heading of VIII xi.: De nimio amore Romae circa Caesarem.

^{page 416 note 1} This is a well-known error in MSS. and is usually, though not always, occasioned by the similarity of the two neighbouring words. An instance in L is victorem for videret in VIII ii. 2: Victorem modo cum victorem Histri. Both P and Q have videret.

^{page 416 note 2} L has preserved IX x., an epigram omitted by PQ, also (likef) malaudit. for male audit in II lvi. 1, a strange form which PQ (very pardonably) omit. The Ca archetype had the same form, which must not be too hastily discarded. It indicates that the two words had coalesced into a ‘word group,’ like mal(e)factum.

^{page 416 note 3} The minuscule form of i which is scarcely distinguishable from minuscule l is the ‘long’ or ‘tall’ form, used by careful scribes only (or mainly) in two circumstances, (1) for consonantal i (j), e.g. aio, (2) for initial i, especially before n. From the first has come our letter j; from the second, if I am not mistaken, the early Irish expression of the preposition in, a preposition which happens to have the same form in Latin, Irish, English, German and other languages. For this preposition Irish scribes often write hi; and linguists have laboured at an explanation of the initial h. I believe the sign to be as conventional as our old-fashioned ‘ye’ for ‘the.’ This was never pronounced with y, but merely written with a sign that resembled the y-siga. No more was the Irish preposition ever pronounced with h.

^{page 417 note 1} The mysterious item in the Geneva Papyrus published by M. Nicole ad hormos confodiendos, a military task, suggested to me the possibility of a Latin loanword from the Greek, hormus, with the sense of ‘trench, chain of trenches.’ But now that L provides hortus, we are absolved from the necessity of explaining ‘obscurum per obscurius.’ A slight variation in the form of early minuscule a makes the letter look like co. In X xii. 9 is the cognoscendus of LPQF a mere scribe's error for agnoscendus, or is it the actual reading preferred by Gennadius? This epigram is not included in the A^A Anthology, but the C^A archetype had adnoscendus. In XIII lxv. 2 the variants lautorum condere and lautorum mandere may have come not from divergent ancient recensions but from the mere blunder of a mediaeval scribe.

^{page 417 note 2} In the B^A family we find this order: I. Epist., ad Cat., (om. i.–ii.), iii.–xiv.; xlviii. 2–ciii. 2; xv.–xli. 3 (om. xli. 4–xlviii. 1); IV xxiv. 2-lxix. 1; 1 ciii. 3–IV xxiv. 1; IV lxix. 2, etc. An account of the causes of this arrangement I hope to give on another occasion.

^{page 417 note 3} Neither of the two MSS. of the B^A family now in the Medicean Library, F and f, suit the description. In a copy of the Paris quarto edition of 1680, at present in the Leyden Library (760 D 5), but formerly in the possession of P. Burmann, there are marginal collations, copied by Burmann ‘ex J. Goyesi codice (i.e. printed edition with written marginalia) qui eum a Beverlando acceperat.’ The manuscript, known asf, a paper codex of the fifteenth century, is the ‘ Florentinus,’ whose readings are recorded by Burmann (e.g. I iii.1 ‘latebras, sed in marg. al. tabernas’). Like F it has culcitra in both passages of the Apophoreta (viz. XIV cxix. 2 and clxii. 1). On the fly-leaf of Burmann's volume there is an entry regarding the Florence MS. twice mentioned by Politian: ‘ Codex Florentinus. Politian. VII Epist. ult. De hoc et aliis vid. Miscell. cap. 29’; but this entry does not give the shadow of a proof that Beverland identified Politian's MS. with f. In the margin at I xix. 2 he characterizes the latter as ‘vet. cod. chartaceus Biblioth. Medicaeae,’ a description which exactly suits f.

^{page 417 note 4} The third transcriber is especially prone to this error. The title-heading of XI xlvi. ‘Ad meuium qui non,’ etc., he has so copied as to make the heading look exactly like a line of poetry and to give the two epigrams xlv. and xlvi. the appearance of one long epigram. One can see the reason of the numerous. cases of ‘conflation’ of neighbouring epigrams in our MSS. of Martial. The first scribe has written the first half of the long heading of V xxxix. in minuscules. But in this he was faithfully copying the archetype, for PQf have precisely the same. If we compare such mistakes with the mistake (mentioned below) made by L in IV xliii. 11, the inference is natural, that the title-headings in the archetype were regularly in minuscules, distinguished from the text by colouring.

^{page 418 note 1} Another case where L gives us the exact reproduction of what was in the archetype is, I think, VI iii. 6 where nebit is the true reading. P has nibit, Q has mbit, f has nubit. In L we find nubit with the second limb of u cancelled by a dot above and below. I fancy the archetype had nubit with an obscure correction of the word to nebit.

^{page 418 note 2} Beside the usual contraction of omnes, viz. oēs, we find occasionally in L ōs. Does this come from the original of L ?

^{page 419 note 1} For instances, see Marx ad Herenn. prolegg. p. 37. An example in L is II lxvi. 6, where the archetype had Tangat et sanum (for isanum probably). L reads Et sanum tangat, Q Tangat et insanum. P does not alter.

^{page 420 note 1} The discovery of L and Prof. Maleyn's recently published account of the readings of P make a continuation of my articles on ‘A Supplement to the App. Crit. of Martial’ unnecessary. I take this opportunity of expressing my thanks to Prof. Fried-laender, Father Ehrle, and Prof. Alfred Monaci for their kind assistance in the compilation of this Supplement.