At the height of a life of studies dedicated to the work of the Latin writer Aulus Gellius, H.-S. has published a new critical edition of the Noctes Atticae and a volume of Gelliana, in which he discusses many passages where he diverges from previous scholars (in the edition marked with *) and in particular from P.K. Marshall's edition (1968).

The text of the Noctes Atticae was transmitted in two separate parts: the first probably contained Books 1–8, the second Books 9–20. This fact caused the loss of Book 8: only the lemmas and fragments remain, reconstructed through Macrobius’ Saturnalia and through Nonius Marcellus. As for the first part (Books 1–7), H.-S. bases his edition on these manuscripts: V (BAV, Vat. lat. 3452, s. XII), C (Cambridge, Clare College 26, s. XIII), P (Paris, BnF lat. 5765, s. XII), R (Leiden, Bibl. Univ. Gronovius 21, s. XII). H.-S. indicates with α the agreement of VC, with ξ that of αPRD (for D, see below), with Ξ that of ξς (ς = lectio vulgata recentiorum). As regards Books 1–4 (especially 1–2), H.-S. adds BAV, Palatinus lat. 24 (A), a palimpsest of the beginning of the third century (fourth century, according to Marshall): the only witness of Gellius before bipartition (Ξ). It preserves a short sequence of lemmata – 17.21 (vi)ri floruerint – 18.2 certa(tionibus) – placed after the praefatio (H.-S. publishes these lemmata in this position and repeats them in a smaller font at the end of each chapter) and a few fragments of Books 1–4. The superiority of A is often evident, for example at 1.3.19 ignoravit A : ignovit Cς : ignovit id V : ignovit it PR; 1.3.20 deieraret Nettleship 1884 : deiraret A : deiuraret Ξ (Gelliana, p. 14); 1.3.27 remissoque A : permissoque Ξ; 2.6.5 rapsatur A : raptatur ΞΦ, Macrob. Sat. 6.7.8; 2.8.8 coniunctionem A : conclusionem Ξ (Gelliana, p. 35); 2.23.21 consarcientis A : consarcinantis Ξ; 2.27.5 prodigendos A, recc. : producendos Ξ; 2.28.1 tremores A : mores VPR : motus CD ^vlς; 2.29.9 necessum A : necesse ΞΦΓ; 3.13.3 quod A : om. Ξ : quos Hertz; 4.1.12 cuiusmodi A : quid Ξ (quod P ^ac).

Even A has errors, later corrected, for example at 1.3.29 derectum] decretum A; 1.4.6 videbitur] videtur A. In the case of 2.29.10 igitur] inquit A, H.-S. (Gelliana, p. 47) says he follows Marache (1967), who prefers igitur, because after the initial inquit by the narrator (in this case, the fabulist Aesop) the verb is used by Gellius only in the characters’ speeches.

Some passages in which H.-S. uses A make it possible to establish some errors in the archetype (Ω): for example at 1.6.8 Îs demum Hertz post Vrsinum (his demum) : isdem Ω; 2.28.1 incompertum Fr. Skutsch : compertum Ω; 4.1.11 praemandere Erasmus Adag. 2.10.33 : praemandare Ω.

These are less certain: 1.3.30 certum Damsté 1919: 290 : cauturum A : cautum Ξ Marshall; 1.22.11 indoctum cod. Ambr. I 129 inf., fol. 352^v (s. xvi^ex), Falster 1721b: 140, Holford-Strevens 1994: 480 : indictum Ω (indicium B^U) Marshall : cf. 19.8.12 indocte et inscite.

At 2.7 Gellius discusses a moral philosophy problem: must we obey all of our father's orders? Of the three opinions proposed, the author prefers the media sententia, according to which sometimes one must obey, sometimes not. In cases where we do not have to obey Gellius says that orders must be rejected leniter et verecunde ac sine detestatione nimia sineque obprobratione acerba reprehensionis. At 2.7.13 H.-S. corrects acerba of Ω with acerbae and explains (Gelliana, p. 35) that ‘all obprobratio is acerba, but not all reprehensio’. In short, the adjective acerba referring to obprobratio would be a pleonasm. But in the lemma of 3.5 Gellius says that certain flaccid and voluptuous behaviours of a man were reproached (obprobrata) by the philosopher Arcesilaus acerbe simul et festiviter. In this case Gellius says that the philosopher has reproached humorously, yet acerbe, i.e. severely. From this we deduce that obprobratio (‘reproach’) may also be not acerba; therefore, H.-S.'s correction acerbae does not seem necessary. (For obprobratio + adjective in Gellius, see 12.12.4: tum Cicero inopinata obprobratione permotus.)

In his edition H.-S. uses, for both parts, three humanistic manuscripts: D (Göttingen, Staatliche Bibliothek, philol. 162, s. XV), w (BAV, Vat. lat. 1532, s. XV), B^U (the so-called Buslidianus, lost, written not before 1469). According to H.-S. B^U dates from the end of the fifteenth century (Praemonenda, p. xxxiv): owned and used by Hieronymus Busleiden and other humanists (as Willem Canter in his Novarum lectionum libri quattuor, Basileae 1564), its readings were preserved above all by Ludovico Carrion (1585). Carrion's bad reputation – he was considered a forger and a thief – inevitably raises doubts about the authenticity of B^U's readings, which mainly concern the first book (but also others, such as 10 and 18). However, excluding the most obvious interpolations, B^U's readings are often correct (e.g. 1.16.11 milli B^U et Macrobiani aliquot (1.5.7) : mille Ξ; 1.21.4 lapidem B^U : lapideum Ξ). Therefore, we believe that H.-S. was right to include also some additions rejected by previous editors, for example at 1.1.3. quanta – conveniret B^U : om. Ξ (Gelliana, p. 10): according to H.-S. this remedies the fact that modificatus est is objectless; moreover, the presence of longinquitas in the sense of length would be a guarantee of authenticity: few interpolators would have thought of it ‘when longitudo lay at hand’ (Gelliana, p. 10).

B^U is particularly valuable because it is the only witness of 18.9.1–8 nam et sequo, missing in the MS tradition. This is a very problematic point, studied from Canter onwards and also by H.-S. in a significant way. At 18.9.1, in the Cato quotation fr. 134 CSbC, the manuscripts have: sed si omnia dolo fecit, omnia avaritiae atque pecuniae causa fecit, eiusmodi scelera nefaria. After nefaria H.-S. completes very well with <neque insecenda>. Gellius quotes a fragment of Cato's oration Contra Thermum to discuss the meaning of the word insecenda (from inseco, an archaic verb with which Livius Andronicus translated the Homeric ἐννέπω), but the quotation lacks the gerundive with attributive function that must obviously be integrated. This is precisely what H.-S. successfully does. The placement of neque insecenda after nefaria is convincing: Thermus’ wicked actions are ‘unspeakable’, ‘not to be told’, and just for this reason they should be told. H.-S. follows F. Skutsch (in Hosius’ edition, Leipzig 1903, Gelliana, p. 152), who wrote atque insecenda. The proposal neque insecenda by H.-S. emphasises well the rhetorical aspect of the sentence. H.-S. holds in great esteem D, a humanistic manuscript, because it transmits excellent readings conjectured by subsequent scholars (list p. xix) and often agrees with the oldest manuscripts. But, as H.-S. admits, D has ‘infinite’ omissions of single words and phrases: its continuous mention in the apparatus often seems heavy if not useless. H.-S. is more balanced in the use of w (which derives from V for the first part and from other Vatican recentiores for the second), mentioned only when it transmits singular readings of a certain importance, for example at 2.10.4 aeditumi Heraeus, cf. 12.10 : edituni w : aeditui VPRD Marshall : aditur C. For the humanistic manuscripts (ς), which derive from a fourteenth-century manuscript (pp. xxviii–xxxi), H.-S. uses the sigla of Hertz's (1886) and Cavazza's (1985–1999) editions. To justify the use, although very limited, of these manuscripts, H.-S. quotes G. Pasquali's famous words, recentiores non deteriores (p. xxx). In fact, in some circumstances ς restores good readings, for example at 1.4.8 enodabat diiudicabatque ς : enum|abatque A ^uv : enudabat deiudicabatque αPR; 4.2.1 servorum ς : scriptorum ξ.

At 6.20.6, giving examples of the pleasant harmony of alphabetic sounds, Gellius quotes Catullus 27.1–4. In line 4 H.-S. prefers ebria acino ebriosioris. That is, he agrees with Haupt's ebria emendation (ebriose VP of Gellius and Catullus’ MSS), but rejects acina of Parthenius (in his Catullus edition, 1485), in favour of acino, a reading transmitted by Gellius’ recentiores (V has instead ac in and P ac me) and Catullus’ potiores. Gellius, however, seems to prefer ebria acina: according to him, Catullus could have said ebrio (therefore acino) because the ordinary form is acinum, of neuter gender, but, as he liked the Homeric hiatus, he wrote ebria, also because it was in harmony with the following a of acina. H.-S. (Gelliana, p. 86) says that a feminine acina was not in use before the late empire and that Catullus used acinus or acinum, changing gender but not declension. It is an eminently Catullian dispute: perhaps H.-S. would be right to prefer ebria acino editing Catullus’ poems, but Gellius read ebria acina, a reading he himself defends: ‘ebriam’ dixit propter insequentis ‘A’ litterae concentum. Discussing this passage and quoting, together with Gellius, another ancient reader of Catullus, i.e. Apuleius, Gaisser says that ‘it is based on an aesthetic argument … and it reminds us that even at this early period there were textual variations and disputed readings in Catullus’ (J.H. Gaisser, Catullus and his Renaissance Readers [1993], p. 13). Therefore, it would be better to print, like Marshall, ebria acina.

In establishing the text, H.-S. sometimes considers epitomes (sec. XIV–XV, pp. xxxi–xxxii: agreement is ε⁺), as at 1.15.9 bibendi ε⁺, recc. : vivendi Ξ.

Florilegia and excerpts complete the manuscript tradition (pp. xxxvii–xl). The Valerio-Gellianum Florilegium (Φ) dates back to the twelfth century and is preserved in twelfth- to fifteenth-century witnesses; the Florilegium written by Guglielmo of Malmesbury († about 1143, φ) is considered to be of lesser value and is preserved in two manuscripts coeval with the author (β is the agreement of Φφ). Among the excerpts, those reproduced by Guglielmo of Malmesbury in his Polyhistor and by Radulfus de Diceto (late twelfth century) in his Abbreviationes Chronicorum are particularly important.

For Books 9–20 the agreement of the principal manuscripts is ψ, while Ψ is that of ψ with recentiores. H.-S. (pp. xx–xxviii) follows Hosius’ classification (Leipzig 1903), identifying two families, γ and δ. The γ MSS are: O (BAV, Reg. lat. 597, sec. IX: O^L are the readings copied by Lupo of Ferrières from another manuscript), X (Leiden, Vossianus lat. F112, sec. X), Π (BAV, Reg. lat. 1646, written in 1170), G (BnF lat. 13038, sec. XII), ν (BAV, Vat. Lat. 3452, sec. XII), N (Florence, BNCF J. 4, 26, sec. XVⁱⁿ, copied by Niccolò Niccoli). From a common copy κ, Gv derive; the siglum π represents the agreement of ΠGν. The δ MSS are: B (Berne, Bibl. civ. 404 + Leiden, BPL 1925, s. XII), Q (Paris, BnF lat. 8664, s. XIII¹), Z (Leiden, Vossianus F7, s. XIII¹). To these two families Marshall added a third one represented only by F (Leeuwarden, Tresoar 55 Hs, written in 836). While admitting the importance of F, which sometimes exhibits the correct reading (e.g. 13.23.3 correptim F : correptam γδ), H.-S. regards Marshall's proposal as a weak stemmatic point of view and believes that F is a contaminated manuscript (γ + δ) with a significant number of singular errors (e.g. 19.1.10 simillimo tui] simili motu^sF), which weaken its authority (p. xxvii). Moreover, H.-S. shows mistrust even when he is almost forced to accept an emendation of F, for example at 11.15.2: Gellius discusses the word amorabunda, a neologism coined by Laberius, after ludibunda and errabunda. However, he says that these two words dicitur ludens et ridens et errans. It is evident that ridens does not have any correspondence. The margin of F has et ridibunda, which is necessary, symmetrically, to et ridens. The anonymous corrector acted on the basis of what Gellius wrote immediately after: 11.15.3 ridens et ridibunda, et ridibunda. The same reasoning probably inspired Andrea Bussi (in the editio princeps, 1469), who in a later passage (11.15.5) added et ridibundo ridens in a sentence in which all MSS have a ludibundo ludens et errabundo errans. H.-S. (Gelliana, p. 113) rightly doubts that both the intervention of F and of Bussi, despite their concinnitas, are only editorial corrections.

In his edition H.-S. has reviewed the vast manuscript tradition and shows that he is perfectly familiar with the immense bibliography of Gellius’ work.

It is impossible in a review to discuss all passages in which H.-S. has modified the transmitted text, using with ἀκρίβεια other scholars’ conjectures. Only in the first three chapters of Book 11 we emphasise words such as vinosus by Meursius (11.2.2: vitiosus Ψ), intercutitus by Bergk (11.2.2: inritus Ψ), summam by Bongars (11.3.4: suam Ψ). In many cases H.-S. proves he is right; in a few others it is legitimate to raise some doubts, as at 12.1.2. One day, Gellius writes, it was announced to Favorinus that the wife of one of his disciples had just given birth to a child: manuscripts hesitate between puerum (X ^pcBZQ) and puerperam (FOπN), but H.-S. accepts the conjecture puerperium, which R. Knöbl has suggested to him privately: Eamus – inquit – et puerperium visum et patri gratulatum. Gellius uses the word puerperium shortly after, but with the meaning of ‘labour’. H.-S. gets round this obstacle (Gelliana, pp. 114–15) with other examples of change in meaning (also in Greek authors) and with the fact that in the same chapter Gellius uses partus both with the meaning of ‘childbirth’ (12.1.8 and 13) and with the meaning of ‘offspring’ (12.1.21). But at 12.1.21 Gellius writes partus suos, and the possessive has an important function to avoid the generic character or ambiguity of the word. The other argument in favour of the conjecture puerperium would be the possibility that in other authors this word means ‘offspring’, ‘child’. In support of this possibility, as Knöbl suggests, H.-S. quotes Sen. Contr. 9.5.1, Plin. N.H. 7.48, 49 and 18.282, Apul. Plat. 2.6 (and also Stat. Theb. 4.280). In Seneca the term is used in the plural (puerperia) and in a generic way as well as in Statius. In Pliny puerperia is used in the sense of ‘fetuses’ (7.48), septem mensum edito puerperio indicates a seven-month-premature childbirth (7.49), and, finally, at 18.282, the word means fruits of the earth. No passage among these can be evidence of the use of the word puerperium in the sense of a ‘a child’, a ‘precise child’. There are also three other generic occurrences: one is internal (NA 10.2.2), and two are not quoted by H.-S., i.e. Varr. Rust. 2.1.26 and Tac. Ann. 12.6.1. We believe it would have been more prudent to accept puerum, well attested, and to quote Knöbl's conjecture in the apparatus. H.-S.'s emendations are not a few, and they are usually excellent. Some of these are in the text, such as dissimile diutius (1.6.8, Gelliana, pp. 17–18), conplusculei (17.2.15, Gelliana, p. 147), dicendi (20.11(10).9, Gelliana, p. 164); others are in the apparatus criticus, due to some corrupt passages marked with ‘crux desperationis’, such as deprecatione (2.6.3 ‘de retractatione’ VR, ‘detestatione’ C Macrob. 6.7.5, Gelliana, p. 33), Caesa <ris Augusti tempo> ribus or <C.> Caesa <ris tempo> ribus, with allusion to Caligula (6.8.2 ‘Caesaribus’ codd., Gelliana, p. 82).

H.-S.'s edition, a true gift to Philology, is destined to be the critical reference text of Gellius’ Noctes Atticae for a long time to come.