There are cases in plenty elsewhere in the Letters to show that B is not many removes from the scriptura continua of some majuscule hand. In the section included in Π, apart from the general tightness of the writing, which led to the later insertion of strokes between many of the words,[35] we note these special indications of a parent manuscript in majuscules. In 61, 10 me autem], B started to write mea and then corrected it. 64, 19 praeceptori a quo] praeceptoria quo B, (m. 1) F. If B or its parent manuscript copied Π directly, the mistake would be especially easy, for praeceptoria ends the line in Π. 64, 25 integra re]. After integra, a letter is erased in B; the copyist, it would seem, first mistook integra re for one word.

Other instances showing a close connection between B and Π are as follows: 62, 23 unice] Π has by the first hand inuice, the second hand writing u above i, and a vertical stroke above u. In BF, uince, the reading of the first hand, is changed by the second to unice; this second hand, Professor Merrill informs me, seems to be that of a writer in the same scriptorium as the first. The error in BF might, of course, be due to copying an original in minuscules, but it might also be due to the curious state of affairs in Π. 65, 24 fungerer]. In Π the final r is written, somewhat indistinctly, above the line. B has fungerer corrected by the second hand from fungeret (?), which may be due to a misunderstanding of Π. 66, 2 avunculi] auonculi Π (o in ras.) B. This form might perhaps be read; F has emended it out, and no other manuscript has it. 65, 7 desino, inquam, patres conscripti, putare]. Here the relation of BF to Π seems particularly close. Π, like MVDoxa, has the abbreviation p.c. On a clearly written page, the error of reputare (BF) for p.c. putare is not a specially likely one to make. But in the blur at the bottom of fol. 52v, a page on the flesh side of the parchment, the combination might readily be mistaken for reputare.

Another curious bit of testimony appears at the beginning of the third book. The scribe of B[36] wrote the words nescio—apud in rustic capitals, occupying therewith the first line and about a third of the second. This is not effective calligraphy. It would appear that he is reproducing, as is his habit, exactly what he found in his original. That original might have had one full line, or two lines, of majuscules, perhaps, following pretty closely the lines in Π, which has the same amount of text, plus the first three letters of spurinnam, in the first two lines. If B had Π before him, there is nothing to explain his most unusual procedure. His original, therefore, is not Π but an intervening copy, which he is transcribing with an utter indifference to aesthetic effect and with a laudable, if painful, desire for accuracy. This trait, obvious in B’s work throughout, is perhaps nowhere more strikingly exhibited than here.

[Π the direct ancestor of BF with probably a copy intervening] If Π is the direct ancestor of BF, these manuscripts should contain no good readings not found in Π, unless their writers could arrive at such readings by easy emendation or unless there is contamination with some other source. From what we know of the text of BF in general, the latter supposition may at once be ruled out. There are but three cases to consider, two of which may be readily disposed of: 64, 3 proferenda] conferenda BF conferanda Π; 64, 4 conprobasse] (comp.) BF comprouasse Π. These are simple slips, which a scribe might almost unconsciously correct as he wrote. The remaining error (63, 28 sibi to si) is not difficult to emend when one considers the entire sentence: quibus omnibus ita demum similis adolescet, si imbutus honestis artibus fuerit, quas, etc. It is less probable, however, that B with Π before him should correct it as he wrote than, as we have already surmised, that a minuscule copy intervened between Π and B, in which the letters bi were deleted by some careful reviser. Two other passages tend to confirm this assumption of an intermediate copy. In 65, 6 (tum optime libertati venia obsequio praeparatur), B has optimae, a false alteration induced perhaps by the following libertati. In Π, optime stands at the end of the line. The scribe of B, had he not found libertati immediately adjacent, would not so readily be tempted to emend; still, we should not make too much of this instance, as B has a rather pronounced tendency to write ae for e. A more certain case is 66, 7 fungar indicis] fungarindicis ex fungari dicis B; here the error is easier to derive from an original in minuscules in which in was abbreviated with a stroke above the i. There is abundant evidence elsewhere in the Letters that the immediate ancestor of BF was written in minuscules; I need not elaborate this point. Our present consideration is that apart from the three instances of simple emendation just discussed, there is no good reading of B or F in the portion of text contained in Π that may not be found, by either the first or the second hand, in Π.[37]

We may now examine a most important bit of testimony to the close connection existing between BF and Π. B alone of all manuscripts hitherto known is provided with indices of the Letters, one for each book, which give the names of the correspondents and the opening words of each letter. Now Π, by good luck, preserves the end of Book II, the beginning of Book III, and between them the index for Book III. Dr. F. E. Robbins, in a careful article on B and F, and one on the tables of contents in B,[38] concluded that P did not contain the indices which are preserved in B, and that these were compiled in some ancestor of B, perhaps in the eighth century. Here they are, in the Morgan fragment, which takes us back two centuries farther into the past. A comparison of the index in Π shows indubitably a close kinship with B. A glance at plates [XIII] and [XIV] indicates, first of all, that the copy B, here as in the text of the Letters, is not many removes from scriptura continua. Moreover, the lists are drawn up on the same principle; the nomen and cognomen but not the praenomen of the correspondent being given, and exactly the same amount of text quoted at the beginning of each letter. The incipit of III, xvi (ad nepotem—adnotasse uideor fatadictaq·) is an addition in Π, and the lemma is longer than usual, as though the original title had been omitted in the manuscript which Π was copying and the corrector of Π had substituted a title of his own making.[39] It reappears in B, with the easy emendation of facta from fata. The only other case in the indices of a right reading in B that is not in Π is in the title of III, viii: ad sueton tranque Π Adsu&on tranqui. B. In both these instances the scribe of B needed no external help in correcting the simple error. Far more significant is the coincidence of B and Π in very curious mistakes, as the address of III, iii (ad caerelliae hispullae for ad corelliam hispullam) and the lemma of III, viii (facis adprocetera for facis pro cetera). ΠBF agree in omitting suae (III, iii) and suo (III, iv), but in retaining the pronominal adjectives in the other addresses preserved in Π. The same unusual suspensions occur in Π and B, as ad sueton tranque (tranqui B); ad uestric spurinn·; ad silium procul.[40] In the first of these cases, the parent of Π evidently had tranq·, which Π falsely enlarges to tranque; this form and not tranq· is the basis of B’s correction—a semi-successful correction—tranqui. This, then, is another sign that B depends directly on Π. Further, B omits one symbol of abbreviation which Π has (possum iam perscri{-b}), the lemma of the ninth letter), and in the lemma of the tenth neither manuscript preserves the symbol (composuisse me quaed). In the first of these cases, it will be observed, B has a very long i in perscrib.[41] This long i is not a feature of the script of B, nor is there any provocation for it in the way in which the word is written in Π. This detail, therefore, may be added to the indications that a copy in minuscules intervened between B and Π; the curious i, faithfully reproduced, as usual, by B, may have occurred in such a copy.

These details prove an intimate relation between Π and BF, and fit the supposition that B and F are direct descendants of Π. This may be strengthened by another consideration. If Π and B independently copy the same source, they inevitably make independent errors, however careful their work. Π should contain, then, a certain number of errors not in B. As we have found only three such cases in 12 pages, or 324 lines, and as in all these three the right reading in B could readily have been due to emendation on the part of the scribe of B or of a copy between Π and B, we have acquired negative evidence of an impressive kind. It is distinctly harder to believe that the two texts derive independently from a common source. Show us the significant errors of Π not in B, and we will accept the existence of that common source; otherwise the appropriate supposition is that B descends directly from its elder relative Π. It is not necessary to prove by an examination of readings that Π is not copied from B; the dates of the two scripts settle that matter at the start. Supposing, however, for the moment, that Π and B were of the same age, we could readily prove that the former is not copied from the latter. For B contains a significant collection of errors which are not present in Π. Six slight mistakes were made by the first hand and corrected by it, three more were corrected by the second hand, and twelve were left uncorrected. Some of these are trivial slips that a scribe copying B might emend on his own initiative, or perhaps by a lucky mistake. Such are 64, 26 iudicium] indicium B; 64, 29 Caecili] caecilii B; 65, 13 neglegere] neglere B. But intelligent pondering must precede the emendation of praeceptoria quo into praeceptori a quo (64, 19), of beaticis into Baeticis (65, 15), and of optimae into optime (65, 26), while it would take a Madvig to remedy the corruptions in 63, 9 (praestatam ad me) and 65,7 (reputare into patres conscripti putare). These are the sort of errors which if found in Π would furnish incontrovertible proof that a manuscript not containing them was independent of Π; but there is no such evidence of independence in the case of B. Our case is strengthened by the consideration that various of the errors in B may well be traced to idiosyncrasies of Π, not merely to its scriptura continua, a source of misunderstanding that any majuscule would present, but to the fading of the writing on the flesh side of the pages in Π, and to the possibility that some of the corrections of the second hand may be the private inventions of that hand.[42] We are hampered, of course, by the comparatively small amount of matter in Π, nor are we absolutely certain that this is characteristic of the entire manuscript of which it was once a part. But my reasoning is correct, I believe, for the material at our disposal.

[The probable stemma] Our tentative stemma thus far, then, is No. 1 below, not No. 2 and not No. 3.