The 4th century Latin dictionary by Nonius Marcellus is our main source for the fragments of lost Latin literature from the Roman republic — works like Accius, the satirist Lucillius, Varro’s Menippean Satires, the Tragedies of Ennius, Sissena and the Historiae of Sallust. The format of the work is a word, a definition, and then one or more quotations to show the usage of the word.
The work was in 20 books, as was traditional for works of grammar. But the books are of very uneven length. In the three volume Teubner edition by W. M. Lindsay from 1903 — still the standard, I believe — volume 1 contains books 1-3; volume 2 contains only book 4, which is vast, and volume 3 contains books 5-20. Book 20 is just a single sheet. The manuscripts reveal that the work was split into these chunks for transmission also.
Three forms of the text have reached us.
The first contains what is known as the ‘pure’ text. This is pretty much untampered with, although subject to the usual perils of transmission. Copying a dictionary composed of short quotes and spotting errors in it is quite a challenge if your Latin is not that good, and Angelo Mai, when he printed the first edition of the text of Cicero’s previously lost De re publica in 1822, described the text as A vertice, ut aiunt, usque ad extremum unguem ulcus est — as ulcerated from top to toe.
The second form of the text is known as the ‘doctored’ text. In some places this is actually more faithful to the original than the corrupted ‘pure’ text. But mostly it has been edited. Some scholar of the Carolingian period revised the text to produce a more readable version, in the interests of those trying to learn Latin. This was a very successful revision, and copies of this version out-number the pure text.
The third form is the ‘extract’ version. The word and definition is included, but the quotations have been omitted in most cases. The result is a glossary, doubtless intended for handier use in monasteries.
All three versions derive from a single archetype, in which a leaf from book 4 had fallen out, and been replaced for safe-keeping immediately after the first leaf of book 1. The transmission is also rather mix-and-match: a single manuscript may use the first form for books 1-3, and the doctored text for book 4.
All the manuscripts are 9th century or later, and all of them, for all three versions, seem to be connected to Tours and the Loire valley in France. In particular the literary activity of Lupus of Ferrieres there in the 9th century seems to be pivotal.
The pure text is represented by the following manuscripts:
L – Leiden, Voss. Lat. F. 73, dated to the start of the 9th century, from Tours.
F – Florence, Lauren. 48.1, 9th century, corrected and annotated by Lupus of Ferrieres.
H – British Library, Harley 2719, 9-10th century. Contains glosses in Breton, so was written in or near Britanny, not far from the Loire. Online.
E – Escorial M.III.14, mid-late 9th century, from Auxerre. The book was at St. Peter’s Ghent during the 11th century.
Gen. – Geneva lat.84, 9th century, from Fulda in Germany, with which Lupus had connections.
B – Berne 83, 9th century, written at Reims in the time of Hincmar.
Cant. – Cambridge University Library Mm.5.22, end of the 9th century, from Bourges.
P – Paris lat. 7667, 10th century, from Fleury.
L contains all three sections of the text, and is a fine and carefully written book made at Tours in the early years of the 9th century, probably while Alcuin was still abbot of St. Martins there. For books 1-3 it is the ancestor of all the other surviving manuscripts above. It incorporates corrections from the doctored and extract families.
The corrections to F are interesting. F3 contains readings and supplements known from no other source, and clearly right. It must be inferred that this corrector had access to another old manuscript — perhaps the archetype of all the manuscripts itself, or a copy taken before the rot had set in.
For book 4, things change. Book 4 of E is descended from book 4 of L, but the best manuscript of this book is Gen. which is NOT descended from book 4 in L, but from some common ancestor. And Gen. was undoubtedly written at Fulda in Lower Germany. There were links between Tours and Fulda, as we can see from the transmission of Apicius and Suetonius, and again we think of Lupus of Ferrieres, whose strong links with Fulda explain why a German manuscript appears in what is otherwise a bunch of manuscripts all written in one area of France. Some of the notes may even be in his hand. We can be reasonably certain that this book was brought from Fulda to the Loire area. Book 4 in B is a cousin of Gen., written rather badly, and the other manuscripts are descended from Gen.
The chunk comprising books 5-20 is different again, with these books in L descended from the archetype, while H, P and E are all cousins of L via one or more now lost intermediaries.
The ‘doctored’ text does not tell us much more about how the text moved around in the Dark Ages. The only complete representative of the whole family is G, Wolfenbuttel Gud.lat. 96. This was written, yes, at Tours between 800-850.
The ‘extract’ family exists in a bunch of manuscripts, and, once again, they are all connected with Tours, Reims, and Auxerre.
Nonius, then, was popular during the 9th century. But he is a difficult author, and after this period he was not copied. Only two medieval book catalogues (St. Vincent, Metz, s. XI, and St.Amand, s.XII) mention a copy. The text did not circulate widely again until the 15th century.