His translator Weiss rates him most highly, and he is doubtless right in considering him modest and fond of learning. Augustine1 calls him “vir elegantissimi eloquii et facundae scientiae,” and Erasmus2 speaks of “Gellii commentariis, quibus nihil fieri potest neque tersius neque eruditius.” He was used by many later writers,3 extensively by Nonius Marcellus and Macrobius.The Manuscripts
Our earliest manuscripts divide the Noctes Atticae into two parts, containing respectively Books i–vii and ix–xx. These were not united in a single codex before the fourteenth or fifteenth century. The eighth book is lost except for the chapter headings and some inconsiderable fragments, a loss which must have occurred between the time of Macrobius, who knew the eighth book, and that of the archetype of our oldest manuscripts; that is, between the fifth and the ninth centuries. That the division of the work was sometimes made after the ninth book is indicated by the epigram of Gaius Aurelius Romanus, which is found in some of the manuscripts at the end of that book; but it would be difficult to account for the loss of the eighth book, if that division had been universal. The manuscripts which contain the whole work are all late, with the exception of the fragmentum Buslidianum. Those which contain the first part, Books i–vii, are the following:
P. Codex Parisinus 5765, of the thirteenth century, in the Bibliothèque Nationale at Paris. It omits i. 1–2. 10 and ends at vii. 4. 3 with the words ictus solis.
R. Codex Lugduno-Batavianus Gronovianus 21, formerly Rottendorfianus. This manuscript is written in various hands, for the most part of the twelfth century. It comes to an end at vi. 20. 6, and it lacks the lemmata.
V. Codex Vaticanus 3452, of the thirteenth century. It begins with the index of chapters, omitting the Preface.
The descent of these manuscripts from a single archetype is shown by the occurrence of the same lacunae (see i. 4. 3 and i. 22. 5 and other examples in Hertz and Hosius), by the same arbitrary additions (iii. 17. 5; v. 18. 9, etc.), and by the same errors (i. 3. 19, 24, 25, etc.). The nature of some of the errors indicates that the archetype of P, R, and V was written in uncials without word-division.
From a different archetype is our oldest manuscript:
A. Palatino-Vaticanus xxiv, a palimpsest, assigned by Hertz to the fifth, by Teuffel (6th ed.) to the sixth, and by Hosius, with a query, to the seventh century. It contains a Latin version of the books of Tobias, Judith, Job and Esther, written over several earlier works: fragments of Livy xci, Cicero pro Fonteio and pro C. Rabirio, Seneca, Lucan and others. Beginning with the 80th folio it contains large parts of Books i–iv of the Noctes Atticae with the addition of a few chapter headings. All the Greek is omitted and a space left for its insertion by another hand. Although carelessly written in “litteris ex quadrata