related to BF, got all ten books from a very ancient manuscript that came down from Paris.  Our simpler stemma indicates the presence of one rather than more than one such manuscript in the vicinity of Paris in the ninth or the tenth century and again in the fifteenth.  This line of argument, which presents not a mathematically absolute demonstration but at least a highly probable concatenation of facts and deductions, warrants the assumption, to be used at any rate as a working hypothesis, that _{Pi}_ is a fragment of the lost Parisinus which contained all the books of Pliny’s Letters.

  [Footnote 43:  See above, p. 44, n. 2.]

  [Footnote 44:  “Zur fruehen Ueberlieferungsgeschichte des
  Briefwechsels zwischen Plinius und Trajan,” in Wiener Studien XXXI
  (1909), p. 258.]

[Footnote 45:  See above, pp. 21, 41.]

[Footnote 46:  See above, p. 22.]

Our stemma, then, becomes,

P (the whole manuscript), of which _{Pi}_ is a part.
|
|
P{1}
/ \
/   \
B    \
F

[Sidenote:  Evidence from the portions of BF outside the text of {Pi}]

We may corroborate this reasoning by evidence drawn from the portions of BF outside the text of _{Pi}_.  We note, above all, a number of omissions in BF that indicate the length of line in some manuscript from which they descend.  This length of line is precisely what we find in _{Pi}_.  Our fragment has lines containing from 23 to 33 letters, very rarely 23, 24, or 33, and most frequently from 27 to 30, the average being 28.4.  These figures tally closely with those given by Professor A.C.  Clark[47] for the Vindobonensis of Livy, a codex not far removed in date from _{Pi}_.  Supposing that _{Pi}_ is a typical section of P—­and after Professor Clark’s studies[48] we may more confidently assume that it is—­P had the same length of line.  The important cases of omission are as follows: 

[Footnote 47:  The Descent of Manuscripts, 1918, p. 16.  Professor Clark counts on two pages chosen at random, 23-31 letters in the line.  My count for _{Pi}_ includes the nine and a third pages on which full lines occur.  If I had taken only foll. 52r, 52v, 53r and 53v, I should have found no lines of 32 or 33 letters.  On the other hand, the first page to which I turned in the Vindobonensis of Livy (133v) has a line of 32 letters, and so has 135v, while 136v has one of 33.  The lines of _{Pi}_ are a shade longer than those of the Vindobonensis, but only a shade.]

[Footnote 48:  Ibidem, pp. vi, 9-18.  There is some danger of pushing Professor Clark’s method too far, particularly when it is applied to New Testament problems.  For a well-considered criticism of the book, see Merrill’s review in the Classical Journal XIV (1919), pp. 395 ff.]

32, 19 atque etiam invisus virtutibus fuerat evasit, reliquit incolumen optimum atque] etiam—­atque om.  BF. P would have the abbreviation for bus in virtutibus and for que in atque.  There would thus be in all 61 letters and dots, or two lines, arranged about as follows: