Reference editions:

For the Commentary on Aeneid 1-4:

Rand, E. K. et al. (edd.).  1946.  Servianorum in Vergilii Carmina Commentariorum, editionis Harvardianae vol. II, quod in Aeneidos libros I et II explanationes continet.  Lancaster, PA. 

Stocker, A. F. et al. (edd.).  1965.  Servianorum in Vergilii Carmina Commentariorum, editionis Harvardianae vol. III, quod in Aeneidos libros III-V explanationes continet.  Oxford. 

For the Commentary on Aeneid  6:

Jeunet-Mancy, E. (ed.).  2012.  Servius, Commentaire sur l’Énéide de Virgile, livre VI.  Paris.

For the Commentary on Aeneid 7-9:

Ramires, G. (ed.).  1996.  Servio, Commento al libro IX dell’Eneide di Virgilio.  Con le aggiunte del cosidetto Servio Danielino.  Bologna. 

Ramires, G. (ed.).  2003.  Servio, Commento al libro VII dell’Eneide di Virgilio.  Con le aggiunte del cosiddetto Servio Danielino.  Bologna. 

Ramires, G. (ed.).  Forthcoming.  Servio, Commento al libro VIII dell’Eneide di Virgilio.  Con le aggiunte del cosiddetto Servio Danielino.

For the Commentary on Aeneid 10-12:

Thilo, G. (ed.).  1878-1887.  Servii grammatici qui feruntur in Vergilii carmina commentarii.  3 vol.  Leipzig.  [repr. Hildeshiem 1961].

The Harvard edition (Books 1-5), that of Jeunet-Mancy (Book 6), and that of Thilo (Books 10-12) have been critically reviewed by G. Ramires on the basis of new collations of the manuscripts and updated in light of other contributions to textual criticism that have been published during the intervening periods. Consider a few examples.  At Aen. 1.12, the Harvard edition has Servius say that the word ‘urbs’ derives from ab orbe, quod antiquae civitates in orbem fiebant, vel ab urbo, parte aratri, quo muri designabantur.  The term ‘urbum’ (or ‘urvum’) is, in fact, the one that Varro also employs to indicate the handle of the plow; however, Codex M (Monacensis 6394, saec. XI) of Servius reads ‘ab uri,’ a reading that Thilo relegated to his apparatus without any comment, but that is confirmed and indeed further explained by Monacensis 15953 (saec. XI), which has ‘a buri,’ obviously the word from which is derived the modern Italian ‘bure,’ the curved rear part of an ancient plow, on which were placed the tiller, the handle, and the plowshare (Vergil uses the term ‘buris’ at Geor. 1.170: continuo in silvis magna vi flexa domatur | in burim et curvi formam accipit ulmus aratri).  In another case, both Jeunet-Mancy and Thilo published the following at Aen. 6.154: sed si quis forte in fluvio pereat nec eius inveniatur cadaver, post centum ei annos ultima persolvuntur.  Codex F (for the Auctus) and Codex L (Leidensis BPL 52, saec. VIII-IX), the oldest and most authoritative testimony for Servius, in place of the final ‘persolvuntur’ read ‘parentatur,’ a technical term for indicating the offering of a funeral sacrifice, which in the present passage seems undoubtedly more appropriate.

Servius’ Commentary on Vergil, and especially that on the Aeneid, has been preserved in a large number of manuscripts that come from the end of the 8^th c. through the 15^th c. (there is an extensive survey of the manuscript tradition in Murgia 1975 and Brugnoli 1990, but cf. also Savage 1932 and 1934, as well as the modern editions of Ramires 1996 and 2003).  The editio princeps seems to be that published at Florence in 1471 by Bernardo Cennini; in the same year, however, Battista Guarini produced an edition at Venice for the Valdarfer press, which was based on the preparations of his father, Guarino Veronese (it was also the object of a pirated edition overseen by Ludovico Carbone, a student of Guarino, on which see Piacente 1987 and Ramires 2008).  Guarino’s edition is also important because the so-called “Italian” additions appear in it for the first time (on this, see the two works of Ramires 2008).  Many editions followed (for a partial survey, see Thomas 1880) until the Parisian edition of 1532, edited by Robert Estienne, which is most noteworthy for the presence of additions that have only recently been attributed correctly to a strand of the manuscript tradition (α, on which see Ramires 1996 and his editions of 1996 and 2003, as well as the specific study that appeared in 2012).  This brings us to Pierre Daniel’s monumental edition, published at Paris in 1600, which for the first time makes note of the important additions that are present in a series of manuscripts and derive from a late-antique commentary that preceded that of Servius, which has long been attributed to Aelius Donatus.  This question will be addressed more fully in the section dedicated to the commentary (see below).  Many other editions followed that of Daniel; some are noteworthy, such as those of Lucius (Basel 1613), Commelin (Leiden 1646), Maswich (Leeuwarden 1717), Burman (Amsterdam 1746), and Lion (Goettingen 1826).  The first modern critical edition was compiled by Georg Thilo and published at Leipzig between 1878-1884.  This remains the only complete edition of Servius.  Indeed, the Harvard edition remains incomplete.  It was started under the authoritative impulse of Rand, but he only published volumes II (Aen. 1-2) and III (Aen. 3-5).  After the definitive studies of Murgia, volume IV (Aen. 6-8) was entrusted to Goold and Marshall, but the two scholars regrettably died without bringing their work to completion.  Of Marshall there remains, inter alia, the monumental work on the Spangenber fragment (Marshall 2000) and perhaps a sketch of the edition of the Commentary on Book 8.  The fifth volume of the Harvard edition (Aen. 9-12) was then entrusted to Murgia, but this great scholar of Servius also died in 2013 before he could publish his edition; this seemed to be on the verge of completion, but was perhaps held back because of the reservations repeatedly expressed by Timpanaro (see the essays republished in his volumes of 1978, 1994, and 2005) and because of the publication of Ramires’ work, cited above.

The editions of Ramires (and to some extent also that of Jeunet-Mancy) are based on the stemma established by Murgia 1975, which proposes two subarchetypes, Δ and Γ.  The first, Δ, can be reconstructed through L (Leiden, Bibliothek der Rijksuniversiteit, BPL 52, saec. VIII-IX); J († Metz, Bibliothèque municipale, Lat. 292, saec. IX); and θ, which itself consists of A (Karlsruhe, Badische Landesbibliothek, Aug. Lat. 116, saec. IX²); O (Oxford, Bodleian Library, Laud. Lat. 117, saec. XI); and S (Sankt Gallen, Lat. 862, saec. XI).  The second subarchetype, Γ, is reconstructed from γ, which itself consists of B (Bern, Burgerbibliothek, Lat. 363, saec. IX²); E (Escorial, Real Biblioteca, Lat. T. II. 17, saec. IX²); Pb (Paris, Bibliothèque nationale, Lat. 16236, saec. X-XI); T (Trent, Biblioteca comunale, Lat. W. 72, saec. IX²); Z (Milan, Biblioteca Ambrosiana, Lat. C 157 inf., saec. XI) and σ, which itself consists of N (Naples, Biblioteca nazionale, Vind. Lat. 5, saec. IX-X); U (Berlin, Staatsbibliothek Preussischer Kulturbesitz, Lat. 4º 219, saec. XII); and W (Wolfenbüttel, Herzog August Bibliothek, Lat. 2091, saec. XIII).  The codices that have been classified as τ were also considered, though these are contaminated, alternately using Δ or Γ as a base text; these include Pa (Paris, Bibliothèque nationale, Lat. 7959, saec. IX); Ps (Paris, Bibliothèque nationale, Lat 7962, saec. IX¹); Q (Florence, Biblioteca Laurenziana, Plut. Lat. 45.14, saec. IX¹); and Sc (Florence, Biblioteca Laurenziana, Plut. Lat. 22.1, saec. IX-X).  There are also those codices of class α, which represent an independent medieval recension of the text, which will be published separately; these include Le (Leiden, Bibliotheek der Rijksuniversiteit, Voss. Lat. F. 25, saec. IX-X); Pc (Paris, Bibliothèque nationale, Lat. 7961, saec. X-XI); and r (Vatican, Biblioteca Apostolica Vaticana, Reg. Lat. 1495 FDG 124r-125v, saec. X-XI).  Many other important codices have been employed occasionally, among which must be mentioned M (Munich, Bayerische Staatsbibliothek, Lat. 6394, saec. XI); Ta (Trier, Stadtbibliothek, Lat. 1086, saec. IX¹); Paris.7963 (Paris, Bibliothèque nationale, Lat. 7693, saec. XII¹); H (Hamburg, Staats- und Universitätsbibliothek, Lat. Scrin. 52, saec. IX); and Mon.15953 (Munich, Bayerische Staatsbibliothek, Lat. 15953, saec. XI).

In Thilo’s monumental edition, Servius’ Commentary on the Aeneid stretches to 1305 pages, but one must consider that Thilo also incorporated (in italics) the so-called Danielis Auctus, which in some cases and in some books is quite extensive.  According to an approximate tally, accounting for the highest density of additions contained in Book 4 (about 50%) and the lowest contained in Books 6 (5%) and 7 (4.5%), while in the remaining books it oscillates between 30-40%, one can say that the “Vulgate” Commentary in Thilo’s edition amounts to slightly less than 900 pages.  The Commentary of Servius, which proceeds according to the schema “lemma + gloss,” does not stop at the interpretation of the Vergilian text, but is rich in grammatical, rhetorical, mythographical, historical, and geographical observations, which are woven together with a full complement of authorial citations; these obviously include Homer (see Scaffai 2006), archaic Latin authors (Plautus, Terence, Ennius, etc.), contemporaries of Vergil (Cicero, Sallust, Horace, Livy), and extend to the “recovery” and “rediscovery” of authors like Lucan and Statius (besides many specific works, see Pellizzari 2003 for an overview).  Because of its breadth and richness, the Commentary of Servius was a point reference for nearly all the late antique and medieval commentaries collections of glosses of other authors (on the relationship with the scholia on Lucan, in particular, see Esposito 2004).  Servius’ Commentary on Vergil, and especially his Commentary on the Aeneid, has been put to good use  (sometimes casually looted) by nearly all readers and critics of Vergil, from Dante (sadly, “Servius” entry in the Enciclopedia Dantesca is still lacking, a void that the studies of Brugnoli 1998, Italia 2008 and 2013, and also Ramires 2010 have tried to fill) to Petrarch (see Feo 1974; Ramires 2002; Fenzi 2011), through to the humanists and other modern critics (see especially the volumes edited by Santini and Stok 2004; Casali and Stok 2008;  Bouquet, Meniel, and Ramires 2011), as well as by authors of encyclopedic and mythographic collections, e.g. Isidore of Seville, the anonymous Vatican Mythographers 1-2, the so-called Vatican Mythographer 3 (perhaps identifiable with Alberic of London, active in the first part of the 12^th c.), and Boccaccio’s Genealogie Deorum Gentilium Libri.”  [G. Ramires; tr. C. L. Caterine].

In the web section Other resources - Modern studies on late antiquity there is a list of all textual variants to the digital edition prepared by G. Ramires; the editor has built the list according to the following criteria:

- The cited variants follow the order of the books from the Praefatio to Book XII^th.

- Every variant is introduced by information that allow a better control of the text.

- Harvard edition (Aen. I-V): the citation includes the page, the verse and the line of Harvard book, in order to make an easy reference to the critical apparatus: e.g. p. 5.1.2 Harv. II means that the locus refers to the second line of the commentary to Aeneid 1, 1, at p. 5.  Harv. II refers to the first and second book of edition, Harv. III to the third, fourth and fifth book

- Jeunet-Mancy (Aen. VI^th): the citation includes page and verse

- Ramires (Aen. VII^th and IX^th): same system used for Harvard edition

- Thilo (Aen. VIII^th, 10^th-12^th): the citation refers to the page, to the verse and to the line following editor's method.

After these suggestions, every textual variant is cited according to the following pattern: first of all come the word or the clause of the base text; then, after a colon, the variant with the abbreviation Th, if the edition comes back to the text of Thilo'edition, Ram. if the chosen text is different and follows manuscript variants ot conjectures made by Ramires or other scholars. [G. Ramires]