enth century, as the previous demonstration has made clear,
our fragment was preceded by 47 leaves that are missing to-day. With
this clue in our possession it can be demonstrated that the manuscript
began with the first book of the _Letters_. We start with the fact that
not all the 47 folios (or 94 pages) which preceded our six leaves were
devoted to the text of the _Letters_. For, from the contents of our six
leaves we know that each book must have been preceded by an index of
addresses and first lines. The indices for Books I and II, if arranged
in general like that of Book III, must have occupied four pages.[3] We
also learn from our fragment that space must be allowed for a colophon
at the end of each book. One page for the colophons of Books I and II is
a reasonable allowance. Accordingly it follows that out of the 94 pages
preceding our fragment 5 were not devoted to text, or in other words
that only 89 pages were thus devoted.

[Footnote 3: The confused arrangement of the indices for Books I and
II in the Codex Bellovacensis may well have been found in the
manuscript of which the Morgan fragment is a part. The space
required for the indices, however, would not have greatly differed
from that taken by the index of Book III in both the Morgan fragment
and the Codex Bellovacensis.]

Now, if we compare pages in our manuscript with pages of a printed text
we find that the average page in our manuscript corresponds to about 19
lines of the Teubner edition of 1912. If we multiply 89 by 19 we get
1691. This number of lines of the size of the Teubner edition should, if
our calculation be correct, contain the text of the _Letters_ preceding
our fragment. The average page of the Teubner edition of 1912 of the
part which interests us contains a little over 29 lines. If we divide
1691 by 29 we get 58.3. Just 58 pages of Teubner text are occupied by
the 47 leaves which preceded our fragment. So close a conformity is
sufficient to prove our point. We have possibly allowed too much space
for indices and colophons, especially if the former covered less ground
for Books I and II than for Book III. Further, owing to the abbreviation
of _que_ and _bus_, and particularly of official titles, we can not
expect a closer agreement.

It is not worth while to attempt a more elaborate calculation. With the
edges matching so nearly, it is obvious that the original manuscript as
known and used in the fifteenth century could not