Chapter 5
Servius

We have already encountered Servius as the author of the commentary on Donatus's Ars that was Pompeius's main source. That is not, however, the work for which he is best known today. Born probably in the late 360s or early 370s, Servius was a teacher at Rome by the 390s. His only writing datable with any security—a brief treatise De centum metris —was probably composed between 400 and 410; its dedication to a clarissimus Albinus (a pupil at the time, as the context shows; probably Caecina Decius Aginatius Albinus, PVR 414) reveals that Servius had by then become connected with one of the most distinguished families of the old capital. His other surviving works include two other concise pamphlets (De finalibus, De metris Horatii ) and the abbreviated version of his commentary on Donatus. The great work that has come down to us is the commentary on Vergil.^[1]

Here too Servius was following in the tracks of Donatus, using his variorum commentary as a major source. Beyond the dedicatory epistle, a vita of Vergil, and the introduction to the Bucolics , Donatus's commentary has not survived. We do, however, have significant extracts thanks to a reader of Vergil—perhaps of the seventh century, perhaps Irish—who incorporated other material into his copy of Servius, including notes from Donatus's commentary.^[2] In the dedication of his work Donatus had said that his compilation was meant to serve as a resource for

[1] See Part II no. 136; for possible evidence of the relative chronology of the commentaries on Vergil and Donatus, see the appendix to Part II no. 125.

[2] The product of the compiler is conventionally termed Servius Auctus or Servius Danielis (after its discoverer, P. Daniel), to be distinguished from the vulgate commentary, which is here taken to be Servius's own. On the relations between the expanded commentary (cited here as DServ.) and the vulgate, see Thilo in Thilo and Hagen, eds., 1, v-lxix; and most recently Goold, "Servius" esp. 102-22; Murgia, Prolegomena 3-6; and Marshall, "Servius." A promising method for reclaiming more of Donatus's commentary has lately been elaborated by Schindel, Figurenlehren . All references to the commentary below are to the vulgate Servius and the Aeneid unless otherwise noted. As in Chap. 4, the extant version of Servius's commentary on Donatus's Ars is cited by page and line of Keil, GL 4.

other grammatici,^[3] and Servius clearly used the work much as Donatus had intended; the learning his predecessor had gathered is excerpted, simplified, and criticized—Servius tends to mention Donatus only to convict him of error—now suppressed, now supplemented.^[4]

It is also clear that Servius's commentary, although a less personal document than the work of Pompeius, is nonetheless the instrument of a teacher.^[5] The commentary remains at the level suitable for pueri as Servius makes his way word by word and line by line through the text, remarking on punctuation, meter, uncertain readings, myth or other Realien , and especially on the language. The last category, in fact, dwarfs all the others, occasioning two notes out of every three. Only one note in seven, by contrast, is concerned with the broader mythical, historical, and literary background of the poetry, and of this small minority only another small proportion amounts to more than perfunctory references or glosses.^[6] The disproportion is a sign of the emphasis that the late-antique grammatici placed on linguistic instruction, which continued well beyond the study of the ars . It reminds us of the distance that separates a modern commentary, given over to exegesis, from its ancient counterpart, in which exegesis coexists—often uncomfortably, as we shah see—with instruction in a living language.

In the central portion of this chapter, then, we shall try to listen to Servius as his students would have heard him, in order to define the impression of Servius's teaching and of Servius himself that would have been fixed in minds more prepared than the modern to appreciate the nuances of his comments and accept them as fresh. Above all, by placing ourselves in the pupils' position we should be able to experience directly one important element of Servius's personality: the grammarian's sense of his own authority. Servius's conceptions of his task and of his status as a cultural figure remain largely unexpressed. Yet in his commentary

[3] See Chap. 4 n. 85.

[4] Note esp. the suppression of many of the references to republican authors that Donatus had included and the inclusion of citations from the newly fashionable Silver Latin poets. See Lloyd, "Republican Authors," with Kaster, "Macrobius" 257.

[5] See Thomas, Essai 182; Lloyd, "Republican Authors" 326; Levy, "To hexês "; Goold, "Servius" 135.

[6] See Kaster, "Macrobius" 256 n. 109.

an individual and often decidedly quirky turn of mind is demonstrably at work, and Servius's implicit self-image at times so influences his comments that they cannot be understood unless it be taken into account.

To begin, however, we will turn briefly away from the commentary to establish a point of comparison with the voice we will later hear and to open a way into Servius's text. It happens that alone of the grammatici in the period Servius speaks both in his own works and as a character in imaginative literature, in the Saturnalia , Macrobius's recreation of the Roman aristocracy's intellectual life in the saeculum Praetextati . Composed just over a generation after the age it celebrates, the Saturnalia offers an idealized Servius standing head and shoulders above the plebeia grammaticorum cohors (Sat . 1.24.8), the good grammarian demonstrating the moral and intellectual qualities desirable in a man of his profession, a teacher "at once admirable in his learning and attractive in his modesty" (Sat . 1.2.15).^[7] As such, Servius is called upon early in the first book (Sat . 1.4) to deal with the adulescens Avienus, who at this point in the dialogue wears the character of a young man essentially sound, if somewhat obstreperous and unformed.

On listening to a discourse by one of the aristocratic participants, Caecina Albinus, Avienus has been struck by the untoward quality (novitas ) of certain turns of phrase the older and more learned man uses. He is moved to question their legitimacy; in effect, he asks why Caecina has committed two solecisms and a barbarism (respectively, noctu futura for node futura, diecrastini for die crastino , and Saturnaliorum for Saturnalium ). The defense of Caecina is entrusted to the professional, Servius, who explains each of the usages in turn and shows that what Avienus in his ignorance took for novitas was in fact antiquitas , the usage of the ancients. The appeal to antiquity fails to impress the adulescens : Avienus savages the grammarian for using his professional status to encourage a way of speaking that time has rubbed out and cashiered. Avienus calls for current language, praesentia verba , until he is brought to heel by the grave rebuke of the group's most distinguished member, Praetextatus himself (Sat . 1.5).

In the conflict that arises from Servius's correction of Avienus, two details are especially important. One is general, Avienus's insistence upon praesentia verba , which is supposedly antithetical to the grammarian's defense of antiquity: in fact, as we will see shortly, Avienus's demand is rather what one might expect from a pupil of the real Servius. The other detail is specific, the method Servius uses to justify diecrastini (Sat .

[7] For general discussion of Servius's role in the Saturnalia and of the context of the incident described below, see Kaster, "Macrobius" 224ff., esp. 243ff.

1.4.20-27), the last of the controversial expressions treated before Avienus's outburst. As commonly elsewhere in the Saturnalia , the words of the speaker—here, Servius—are drawn from a chapter of Aulus Gellius (10.24); and, as is his practice, Macrobius substantially rearranges and modifies the chapter to suit his purpose. Servius's defense proceeds from the assertion that the doctissimus vir Caecina did not use the expression sine veterum auctoritate . The method of the defense, and so the use of auctoritas as a criterion of correctness, is essentially analogical; that is, Servius adduces no attested use of diecrastini to provide an authoritative precedent, nor does he even claim (as did Gellius) that the form was ever used by the ancients.^[8] Rather, the expression is justified solely and explicitly by analogy with such attested archaic forms as diequinti and dienoni .^[9] The fictional Servius's defense of analogical archaism and the respect for the veterum auctoritas that it implies conform thoroughly to Macrobius's idealized vision of the literary culture: they are in accord both with Servius's role as the good grammarian, the man who guarantees the continuity of the language, and with the more general notion that stamps each page of the dialogue, the belief that the cultural tradition continues as a living presence, influencing and validating every aspect of a mature and learned man's life. At the same time, the defense of antiquity that Macrobius's Servius offers and the regard for auctoritas that analogical archaism implies are directly opposed to the doctrines of the Servius we find in the commentary.

The real Servius's view can be seen in several notes on the Aeneid . Characteristically, the instruction appears early in the first book, so that the student may carry the lesson with him as he proceeds:

1.4 MEMOREM IVNONIS OB IRAM constat multa in auctoribus inveniri per contrarium significantia: pro activis passiva, ut [11.660] "pictis bellantur Amazones armis," pro passivis activa, ut [Georg . 1.185] "populatque ingentem farris acervum," et haec varietas vel potius contrarietas invenitur etiam in aliis partibus orationis . . . et in nomine, ut "memorem Iunonis ob iram"—non "quae meminerat" sed "quae in memoria erat." de his autem haec tantum quae lecta sunt ponimus nec ad eorum exemplum alia formamus.

[8] In a sentence omitted by Macrobius, Gellius does offer the assurance that item simili figura "diecrastini" dicebatur, id erat "crastino die" (NA 10.24.8); the only attested use of diecrastini before Gell. NA 2.29.7 is Plaut. Most . 881, a fact of which Macrobius would almost certainly have been unaware.

[9] Sat . 1.4.20, with 1.4.25-27. The analogical defense is found clearly at Sat . 1.4.25, verum ne de "diecrastini" nihil retulisse videamur, suppetit Caelianum illud ex libro Historiarum secundo , citing a passage that contains diequinti . Symmachus and Praetextatus then follow with quotations offering diequinti and dienoni .

The final sentence warns against imitative extension of the peculiar usages found in the text and conveys the main point of the note.^[10] The principle found there can be compared with the burden of another note, which occurs not long after:

1.26 ALTA MENTE REPOSTVM. . . . "repostum" autem syncope est: unam enim de medio syllabam tulit. sed cum omnes sermones aut integri sint aut pathos babeant, hi qui pathos habent ita ut lecti sunt debent poni, quod etiam Maro fecit: namque et [6.655] "repostos" et [8.274] "porgite" de Ennio transtulit. integris autem et ipsis utimur et eorum exemplo aliis.

The main thrust of the note (hi . . . poni and integris . . . aliis ) moves in the same direction as the comment on Aeneid 1.4. Both notes concern the use and abuse of analogy and the proper relation between analogical formation and auctoritas :^[11] the combined lesson is plainly opposed to the validation Servius gives diecrastini in the Saturnalia , where the one odd expression is justified merely by analogy with similar odd expressions in the texts of literary auctores . Such notes represent specific and limiting applications of the general statement concerning figurative usage found later in the commentary:

5.120 PVBES INPELLVNT figura est, ut [1.212] "pars in frusta secant." et sciendum inter barbarismum et lexin, hoc est, Latinam et perfectam elocutionem, metaplasmum esse, qui in uno sermone fit ratione vitiosus, item inter soloecismum et schema, id est, perfectam sermonum conexionem, figura est, quae fit contextu sermonum ratione vitiosa. ergo metaplasmus et figura media sunt, et discernuntur [sc. from barbarism and solecism, respectively] peritia et imperitia. fiunt autem ad ornatum.

Compared with the definitions found in the grammatical tradition, Servius's note here is distinctive in several details^[12] and can be contrasted

[10] In the final sentence I have read ponimus and formamus with Thilo and the manuscripts of Servius, against posita sunt and formata , the readings of codex C of DServ., printed by the Harvard editors: for the reason, see p. 181 and n. 34.

[11] For a similar formulation, cf. Servius's comment at 1.587. See also Serv. 441.13-15, with Pomp. 263.11-28 (cf. Pomp. 232.2-8 and 12-14, and the still more stringent formulation at 187.15f.); and the elder Pliny ap . Charis. 151.18-25 Barwick, warning against the analogical extension of archaic nominal forms in -es (i.e.,copies , on the model of amicities ) and opposing the rationis via to the veterum licentia .

[12] Esp. in its precise tripartite schematization (lexis, metaplasmus, barbarismus , and schema, figura, soloecismus ) and in the formal categories of lexis and schema . Schema itself is commonly used interchangeably with figura to denote what in ordinary discourse would be considered a solecism: e.g., Don. 658.3H.; Serv. 448.1-7; Pomp. 292.13ff. = Plin. frg. 125 della Casa (above, p. 151f.); cf. also Quintil. Inst . 9.3.2, with Don. 663.5-6H.

with the less precisely worded doctrine in the extant version of his commentary on Donatus's Ars :

quidquid scientes facimus novitatis cupidi, quod tamen idoneorum auctorum firmatur exemplis, figura dicitur. quidquid autem ignorantes ponimus, vitium putatur. (447.8-10)

This last, broad formulation, with its emphasis on novelty—novitas , toned down to the less daring ornatus in the Vergilian commentary, where little good is said about novitas —and with its vague proviso concerning auctoritas (idoneorum auctorum firmatur exemplis ), could perhaps be taken to countenance the kind of analogical argument offered in Macrobius. The comment on Donatus provides a general, liberal alternative to the specific and confining statements found, for example, in the notes on Aeneid 1.4 and 1.26. In the latter places, it seems, we hear the authentic and assertive voice of Servius the teacher,^[13] a voice distinct from that of the good grammarian of the Saturnalia .

There is more at stake here than just another variation in detail between the creation of Macrobius and the author of the commentary.^[14] The two figures understand and value in fundamentally different ways the processes of the language, the authority of the culture that stands behind it, and the status of the grammarian himself. The practice of analogy in the Saturnalia clearly accords with the ideal of cultural continuity developed in the dialogue. In using that approach, one assumes that the forms guaranteed by auctoritas are—to adapt the term Servius used in the note on Aeneid 1.26—as "sound" (integra ) as the forms used in regular speech and thus are as suited to the operations of analogy; through that linguistic exercise one achieves a more intimate and vivid participation with the ancients.

Precisely the opposite is true of the teaching of Servius, for whom auctoritas holds no such guarantees: figurae (or metaplasms, which operate under the same terms) are a large but finite and isolated repository of ancient expressions.^[15] The repository is, above all, controllable; it is not

[13] On Servius's tendency to vary his teaching in the commentary on Vergil, "where he was not as bound to the [grammatical] tradition" as in his observations on Donatus, see the remarks of Wessner, "Lucan" 329.

[14] For such differences in general, see Kaster, "Macrobius" 255ff.

[15] Just how finite and isolated can be gauged from such collections as the Exempla elocutionum of the rhetorician Arusianus Messius (edited most recently by A. della Casa [Milan, 1977]), based on Terence, Vergil, Sallust, and Cicero. Cf. the singularia gleaned from the works of Cato and Cicero in the second century by Statilius Maximus, discussed by Zetzel, "Statilius."

to be extended. Figurae may be used under certain conditions virtually as literary allusions, but at the same time they exemplify what should be avoided as vicious in general practice. In Macrobius, figurae represent a free channel of communication between past and present that the grammarian has modestly and reverently opened; in Servius's commentary, figurae represent a nearly closed door over which he stands guard. The ends of immediacy and participation that the grammarian of the Saturnalia serves in instructing young Avienus are countered in Servius's own teaching by the preservation of distance and control.

The goals of distance and control are themselves partially the result of an institutional quirk of Roman education. Figurae occupied a no-man's-land between the schools of the grammaticus and of the rhetor , falling a bit short of the latter's main interest but a bit beyond the former's central concern, the correct understanding of the parts of speech and their attributes.^[16] This institutional no-man's-land coincided with a no-man's-land of language and method. The ambiguous place of figurae in the structure of formal education conditions the ambiguous function of figurae in the commentary, where they commonly mark the boundary between two opposing ideas (e.g., exegesis vs. prescription, the ancients vs. "us," the language of Vergil vs. correct language) but at the same time leave it porous or vague. So, for example, in the economy of Servius's commentary, figurae mediate between the two main purposes, exegesis and prescription: figurae make intelligible what the author is saying (and often defend his way of saying it) while segregating the author's usage from the grammarian's central lesson of correct speech.^[17] In any given note, one purpose may predominate, but the boundary between the two is never neat; one should perhaps speak not so much of boundaries as of buffer zones. The institutional niche of figurae corresponds to their use as a buffer (compare Servius on Aeneid 5.120, quoted above, where such usages are termed media ): the category figura protects

[16] See the remarks of Schindel, Figurenlehren 12ff. Servius most often uses figura in the sense of the schema grammaticum , a deviation from the loquendi ratio , as defined by Quintil. Inst . 9.3.2; cf. Serv. 448.1-7, opposing the schema in sermone factum (= the figura grammaticalis ) to the schema in sensu factum (= figure of thought, the sphere of the rhetor ). On the place of the parts of speech at the heart of the grammarian's expertise, cf. Chap. 4 pp. 140, 163.

[17] This is grammar's quod liter Iovi, non licet bovi ; cf. Aug. C. Faust . 22.25 (PL 42.417): puer in barbarismo reprehensus, si de Vergilii metaplasmo se vellet defendere, ferulis caederetur . The dispensation is usually but not always extended to Vergil: see Servius at, e.g., 4.355, 8.260, 10.572.

the regular operations of the language against the authority of the text just as it protects the text against the charge of solecism.^[18]

As the goals of protection, distance, and control suggest, the commentary is often a scene of conflict, between the ancients and "ourselves," between different forms and sources of authority, between the deference owed to the author's prestige and the grammarian's domination of the text. Understanding the commentary means in large part understanding how the grammarian controls such conflicts, and understanding that control requires us to appreciate the sense of authority that the grammarian derives from his own institutional niche. As is often pointed out, Servius's approach to the text is one of regulations and categories; and this quasi-bureaucratic treatment of Vergil has done little to endear Servius to modern tastes. But bear in mind that controlling regulations and categories carries a power with it. Servius understands whatever comes before his eyes through the rules his institution provides, and he owes whatever authority he possesses to his command of those rules and to his status in that institution. As we shall see, Servius has so thoroughly internalized those rules and the authority of his position that they are at times combined and expressed unconsciously, in ways that offer unexpected glimpses of Servius's personality and self-image.

Before considering his self-image, however, we must try to understand the basis of Servius's rules and authority and how they are deployed in the commentary. Servius's status as a grammarian, his place in the specialized institution of his profession, involves a specific knowledge, recte loquendi scientia , which is presumed to rest on the natura of the language. Recte loqui means naturaliter loqui : strictly correct usage is natural usage.^[19] Natura provides the raw material of the language, from, say, the quantity of the root vowel of unus or the correct spelling of scribo to the various functions and forms of the parts of speech.^[20] This raw

[18] See n. 17; compare Servius at, e.g., 1.120, on the construction of Ilionei .

[19] Compare the phrase sermo natura . . . integer (implicitly the claim of the grammarians) in the polemics of Arnob. Adv. nat. 1.59: above Chap. 2 p. 85. On the grammarian's presumption of a natural order in language, see most recently Blank, Ancient Philosophy esp. 13, 51 n. 1.

[20] The most venerable text on this subject is preserved by Diom. 439.16-22 = Varro frg. 115 Goetz-Schoell: [Latinitas ] constat . . . . ut adserit Varro, his quattuor, natura analogia consuetudine auctoritate. natura verborum nominumque inmutabilis <<est> nec quicquam aut minus aut plus tradidit nobis quam quod accepit. nam si quis dicat "scrimbo" <<pro eo> quod est "scribo," non analogiae virtute sed naturae ipsius constitutione convincitur. analogia sermonis a natura proditi ordinatio est secundum technicos ; Charis. 62.14-63.9B. offers a parallel version. Ratio and analogia are used interchangeably in the passage; see esp. Diom. 439.16f., with 439.27f. On auctoritas and consuetudo (= usus ), see below. The relation between natura and ratio or analogia here was correctly seen by Barwick, Remmius 183f.: "analogia . . . ist die der natura gegebenen Sprache abgewonnene Gesetzmässigkeit. Daher sind natura und analogia bis zu einem gewissen Grade zwei verschiedene Seiten ein und derselben Sache."

material is subject for the most part to ratio (or analogia ), which systematically orders the data of nature to provide the regulae set down in the ars ; the ars in its turn is the product and property of the grammarian. The nature of the language is thus incorporated in the institution and identified with the grammarian's expertise. The linguistic forces that lie beyond his institutional niche and contradict his expertise are against nature.

Consider, for example, the comments on figurative usage noted above. By definition a deviation from correct usage, figurae are also necessarily a deviation from natural usage, the sermo naturalis.^[21] Inevitably, therefore, the grammarian is as opposed to analogically extending figurative usage as he is to extending any other usage against nature: with Servius's note on Aeneid 1.4, we can compare the following, which warns against back-formation from a form whose natura has been corrupted:

2.195 PERIVRI in verbo "r" non habet; nam "peiuro" dicimus corrupta natura praepositionis. quae res facit errorem, ut aliqui male dicant "peiurus."^[22]

Here, as in the case of figurae , an accommodation must be reached with the corruption already accomplished: Servius is saying, in effect, "This far, but no farther."

Hedged around by the wall of natura , Servius deals from a clearly defined position of strength with the other, unruly forces—auctoritas (literary authority) and usus or consuetudo (ordinary, current usage)—that have an impact on the language. These forces are variously treated in the commentary. For example, auctoritas serves largely as a court of last resort, defining the periphery of permissible usage rather than the core of what is correct;^[23] but auctoritas can also appear, now and then, to govern the language when that serves the grammarian's didactic

[21] So 2.132, figura vs. sermo naturalis . See also 1.5, expanded at 5.467: usage adopted causa metri vs. what one does naturaliter ; 2.60, usurpatum vs. naturale (on this comment see below, p. 182); 7.161, secundum naturam vs. figuratum .

[22] See also the comments at 4.427, against the 1st sing. perf. indic. revulsi as derived from the "unnatural" form (re )vulsus ; and cf. 2.39 (sim. 1.149), on the declension of vulgus .

[23] See above, on 1.4 and 1.26; similarly Pomp. 232.2-8 (with Serv. 437.20-23), 237.35ff., 263.11-28 (with Serv. 441.13-15), 273.3ff.; cf. also Diom. 370.19-23 for reliance on auctores in a case where non est inventa ratio .

purpose.^[24] Much the same is true of usus : what usage has maintained can be a determining factor,^[25] and can even be credited with altering the nature of the language.^[26] But throughout, the grammarian, with his understanding—or rather, definitive control—of natura , stands watch over auctoritas and usus , guarding against the perceived abuse, confusion, and corruption that both produce.^[27]

This intricate and often arbitrary interweaving of natura, usus , and auctoritas is tolerably familiar;^[28] only two tendencies need emphasis here. First, one can reasonably suggest that the authority of the grammarian's own pronouncements would be perceived by his students and by the grammarian himself as dominant and decisive: the grammarian establishes the distinction between "what we read" and "what we say," grants his permission according to his notion of "what we are able to say," determines the propriety of particular usages, and above all issues warnings.^[29] Second, when that authority is blended with the prescriptive

[24] See, e.g., 8.409, 12.587; cf. Servius's comment on Georg . 3.124, and, on Servius's often arbitrary invocation of idonei auctores , see Kaster, "Servius." For auctores or veteres used to confirm a rule reached by way of analogy, cf. Diom. 368.3-11, 375.16-25.

[25] Cf. 2.268, where Servius notes the divisions of night and day according to Varro and concludes, de crepusculo vero, quod est dubia lux, . . . quaeritur, et licet utrique tempori [i.e., twilight or daybreak] possit iungi, usus tamen ut matutino iungamus obtinuit . The meaning "daybreak" authorized by usus according to Servius in fact first appears outside Servius in Latin of the fourth and fifth centuries; see TLL s.v. crepusculum 1175.39ff.

[26] Cf. 5.603, HAC CELEBRATA TENVS tmesis est, "hactenus." et hic sermo, quantum ad artem [i.e., naturam ] special dubs continet partes orationis, ut "hac" pronomen sit, "tenus" praepositio. . . . sed iam usus obtinuit ut pro una parte babeatur. ergo adverbium est: omnis enim pars orationis, cum desierit esse quod est, in adverbium migrat ; for the principle, see above at Chap. 4 n. 16. Here again the connection between the ars and the essence of the language, quod est , is plain.

[27] To select from only the first book a few examples in which auctoritas and usus are found abuti or confundere or corrumpere . For auctoritas , see 1.118, 185 (on the distinction between totus and omnis , never firm, and weakened further in the common speech of late antiquity [see Löfstedt, Late Latin 22]: Servius here urges against the improper usage of the text as a way of undermining a bad habit of common speech; cf. pp. 187-89 below), 334, 590. For usus , see 1.319, 410, 480, 697 (sane sciendum malo errore "cum" et "dum" a Romanis esse confusa —again touching upon an authentic feature of late Latin; see Adams, Text 77). For regula or ratio corrupted by consuetudo or contradicted by the license of the veteres or of current usage, cf. Diom. 348.24f. (cf. 349.6f., 15f.), 365.25f., 398.9, 400.3, 406.11-12.

[28] On how the scholars of the first century B.C. and first century A.D. balanced the components of the language, the general discussion of Barwick, Remmius 203-15, remains fundamental; also valuable, esp. for the Greek background, is Siebenborn, Sprachrichtigkeit esp. 108-15 and, a bit less satisfactory, 151-54, on natura . For an accessible and useful introduction, see Bonner, Education 204-8.

[29] "What we read" vs. "what we say": e.g., 2.487, 3.278, 7.605. Possumus uti or licenter utimur or pro nostro arbitrio utimur : e.g., 1.47, 96, 159, 177, 194, 343, 430, 451, 484; 2.610. Sphere of usage: cf. 1.251, 2.18, 6.79, 10.481. Warnings: cf. 2.513, VETERRIMA usurpatum est. ergo, ut supra diximus, hoc tantum uti si necesse sit licet , with 1.253, HONOS cum secundum artem [i.e., naturam ] dicamus "honor," "arbor," "lepor," plerumque poetae "r" in "s" mutant causa metri. . . . hoc quidem habet ratio: sed ecce in hoc loco etiam sine metri necessitate "honos" dixit. item Sallustius paene ubique "labos" posuit, quem nulla necessitas coegit. melius tamen est servire regulae [i.e., naturae ]. Compare the still stricter position of Pomp. 283.37ff.

purposes of the commentary, Servius's manufacture of the language for his students' benefit can produce observations on Vergil's language that sound absurd to the modern ear, attuned as it is solely to the commentary's exegetic purpose. Neither tendency can be separated from the other, but the effects of the second are more easily seen in individual notes. We can therefore examine this second tendency and identify a few of the strategies Servius used in his instruction, before returning to consider the implications of the first.

We must accustom ourselves to hearing Servius with a student's ear when he says, for example:

10.526 PENITVS DEFOSSA TALENTA. . . . sane melius [i.e., rectius] "infossa" diceret quam "defossa," ad quod est metri necessitate conpulsus.

Or:

11.468 ILICET "confestim," "ilico": quod ne diceret, metri necessitas fecit. nam "ilico" dicimus.

These and other notes invoking the necessity or compulsion of meter (and anyone familiar with the commentary knows how common they are) were not intended, and would not have been understood, as purely or even primarily exegetic. They are not earnest but superficial attempts to judge or explain Vergil's own choices and technique; rather, the force of these observations is directed largely at the student, telling him what he should or should not do. That the words are Vergil's is virtually incidental. Freely paraphrased, these lessons would be understood to mean something like, "Don't get it into your head that you should do what Vergil has done here; your usage should be such only when all other options have been dosed."^[30] The text serves as an instrument; the author, as dummy. Both are exploited to meet Servius's purpose.

[30] Cf. 2.513 and 1.253, cited in the preceding note; and esp. 1.3, p. 187 below. With the paraphrase offered in the text, compare what Pompeius says at, e.g., 269.22ff., ne dicas mihi ergo, "quoniam usus est Vergilius . . .,debeo et ego ita facere." nequaquam licet . Cf. n. 17 above.

The exploitation recurs over a wide range of Servian rhetorical ploys. A similar and, again, essentially negative tactic involves the use of the phrase debuit dicere . For example:

1.16 HIC ILLIVS ARMA figura creberrima adverbium pro adverbio posuit, praesentis loci pro absentis: debuit enim dicere "illic."

Or:

9.467 CINGITVR AMNI "amne" debuit dicere: numquam enim bene [i.e., recte] in "i" exeunt nisi quae communis sunt generis, ut "docilis," "agilis." sed ideo ausus est ita ponere ablativum, quia, ut supra diximus [9.122], apud maiores "hic" et "haec amnis" dicebatur.

Again, debuit dicere is directed more at the student than at the text. Servius is not literally contending at Aeneid 1.16 that Vergil should have said illic ; he is making plain to the students what they should use. At stake is not so much a fault worthy of criticism or demanding correction in Vergil—figura provides the necessary protection against that—but a deviant usage the student should avoid.^[31] Similarly, ausus est in the note on Aeneid 9.467 is not meant to describe Vergil's behavior, for his daring is immediately denied by the explanation that his words simply reflect the usage current apud maiores , of whom he is one. Rather, ausus est is directed at the student, to impress upon him what should be avoided as bold.^[32]Debuit dicere urges against the bad example of the text and has the effective meaning in the commentary of debemus dicere .^[33] Such notes drive

[31] But dearly debuit dicere could be construed as an adverse criticism of the author; cf. especially Aug. C. Faust . 22.25, quoted below, n. 77, and note Servius and DServ. at 1.273: where Servius offers GENTE SVB HECTOREA id est "Troiana." sed debuit dicere "Aeneia." diximus superius [1.235] nomina poetas ex vicino usurpare , explaining and defending as a poetic usage the deviation from what should have been said, the same idea appears in DServ. as GENTE SVB HECTOREA id est "Troiana." sed quidam reprehendunt quod "Hectorea" et non "Aeneia." mos est poetis nomina ex vicinis usurpare ; here quidam reprehendunt is probably a generalizing inference drawn from Servius's debuit dicere by the compiler of DServ. (the last clause is certainly no more than a finicky rewriting of Servius typical of the compiler). Debuit dicere seems to be located, like figura , in a gray area: although the practical or monitory purpose of the phrase dominates in its many appearances in the commentary, it would be wrong to deny out of hand that debuit dicere could connote some criticism of Vergil.

[32] Cf. 2.610 (concerning a point of usage similar to that in 9.467), where each of the three parts of Servius's statement, "tridente" debuit dicere; sed novitatem adfectavit, nulla cogente necessitate , is intended more as a warning for his pupils than as an objective interpretation or evaluation of the verse.

[33] Cf. 1.319, facere non debemus , referring to a Graeca figura .

home their lessons through the use of the third person singular, segregating the author's usage from the Latin that Servius wants to teach.

In these examples, prescription proceeds obliquely, yet nonetheless dearly, as Servius plays his own views off against the text. But prescription is at work in another element of Servius's style, one that is not at all apparent on the surface of his language but is wholly implied in his role. It is a nuance that again requires us to hear Servius with the ear of his pupils and that, not incidentally, adds to the difficulty of teasing apart the strands of natura, usus , and auctoritas in Servius's weave.

We might look again at the notes on Aeneid 1.4 (for its last sentence) and 1.26:

de his autem haec tantum quae lecta sunt ponimus nec ad eorum exemplum alia formamus.

sermones . . . qui pathos habent ita ut lecti sunt debent poni. . . . integris autem et ipsis utimur et eorum exemplo aliis.

As noted earlier, we should accept the readings of Servius's manuscripts at 1.4, ponimus and formamus , against posita sunt and formata , the readings of Servius Danielis, which some editors have imposed on Servius.^[34] The reason for following the manuscripts of Servius is simple. They bear witness to a constant feature of Servius's language, the use of the first-person-plural indicative in a prescriptive sense.^[35] That is, ponimus , unless otherwise qualified, would tend to mean ponere debemus , or formamus to mean formare debemus , or utimur to mean uti debemus —compare the parallel uses of the verbs at 1.4 and 1.26, ponimus . . . formamus and debent poni . . . utimur. This thoroughly natural overtone is unmistakable as soon as one listens to Servius as a teacher of his native language and not simply as a descriptive, objective commentator in the modern vein. The nuance occurs throughout the commentary, as in some of the notes already cited.^[36] It is found most easily, perhaps, in one of Servius's more striking

[34] See above at n. 10. The alteration of Servius's active forms to the passives in DServ. is comparable to the systematic alteration of Servius's formula of cross-reference, ut supra diximus , to the impersonal ut supra dictum est in DServ., revealing the hand of the compiler. Cf. Goold, "Servius" 107-8; Murgia, Prolegomena 100-101.

[35] This nuance was clearly understood by Thilo: see his remarks in Thilo and Hagen, eds., 1, lxxii.

[36] E.g., 11.468, nam "ilico" dicimus , where dicimus plainly serves the purpose for which debuit dicere is used elsewhere. The nuance is of course not peculiar to Servius. A striking example occurs at Pomp. 238.17-19, est verbum quod regit dativum, "maledico tibi." et hoc in usu pessime habemus: nemo dicit "maledixit me ille," sed dicimus "maledixit mihi" tantum modo. The sense of hoc in usu pessime habemus here becomes dear only once one realizes that Pompeius is saying, with notable compression, et hoc in usu pessime habemus: [multi enim "maledixit me ille" dicunt, sed vitiose. nam ] nemo dicit [= debet dicere ] "maledixit me ille ," sed dicimus [= debemus dicere ] "maledixit mihi. " The usage with the accusative does appear to have been vulgar; see TLL 8.164.13ff., and cf. CGL 3.112.19, 49.

pieces of instruction, where he urges the obsolete pronominal form ipsus against the common but irregular ipse :

2.60 HOC IPSVM "ipsum" autem per "m," quia usurpatum est "ipse," et est naturale "ipsus," ut [Ter. Andr. 576] "ipsus mihi Davus." dicimus ergo "ipsus, ipsa, ipsum," ut "doctus, docta, doctum."

Here, as Servius stresses what is regular and natural, the meaning of dicimus slides entirely into the realm of what should be, leaving simple description behind.^[37]

Beyond demonstrating how readily description is subordinated to prescription, this last example deserves attention for a related feature, the nonchalance with which Servius identifies a palpable archaism (ipsus ) with what "we [ought to] say." Or perhaps it is more accurate to say that ipsus would seem to us moderns a palpable archaism; for in Servius's language, both dicimus and naturale effectively deny that ipsus is an archaism at all. Rhetorically, as a means of confirming the lesson, naturale plays the more important role: by associating ipsus with nature, the epithet distinguishes the form from ambivalent usages like figurae and guarantees its simple, regular validity. There is, plainly, a fair amount of room for eccentric judgment in such matters as this, in part because the concept of natura is itself a bit vague about the edges and has its own eccentricities.

So much becomes apparent as soon as one tries to pin the concept down. The natura of the language cannot be defined historically in any straightforward way, as something that once came into being with specific characteristics, some of which have endured through the passage of time while others have become obscured or distorted. As we shall see just below, Servius does not believe that the farther back in time one probes the closer one comes to natura , or that the usage of the ancient authors, the maiores , reveals the language in a pure or more natural state. Yet the natura of the language is not timeless, an abstraction somehow outside history, for it can be affected in and over time: not only can natura be corrupted, and not only can usus change the nature of parts of speech,

[37] An obvious variant of the prescriptive indicative—what might be called the permissive indicative—is found in notes where such phrases as dicimus et . . . (et ) and utrumque dicimus in effect mean licet or possumus dicere et . . . (et ), and licet or possumus utrumque dicere : e.g., 1.484; cf. Serv. 418.33f. and 442.1f., with Pomp. 269.32-34.

but Servius's treatment of archaism even implies at times that in its nature the ancients' language overlaps only partially with his own.^[38] When the shift is imagined as having occurred, or whether it has ceased to occur, is not made dear, although, as we shall see, a primary agent of the change would appear to be the grammarian himself. At this moment, however, the points to be emphasized are these: to the extent that natura inheres in the grammarian's institution, in the form of rules (regulae , the guarantee of what is rectum ), it provides the grammarian with a stable place to stand; and like figurae , archaisms, where they are noted, implicitly involve usages that not only contradict the lesson Servius wishes to teach but also run against nature. In our attempt to gauge Servius's sense of his control of the language, therefore, it is important to understand what he has in mind when he deals with archaism in the opposition between antiqui and nos.

It is evident that when Servius identifies one of Vergil's usages as antiquum , an archaism, he does not mean that it was an archaism in Vergil's time (although it might have been that as well) but that he judges it to be obsolete when tested against his own complex sense of acceptable current usage. Vergil was himself one of the antiqui (maiores, veteres ) and was grouped as such, in a broad stroke characteristic of ancient scholarship, with the classical and preclassical authors; although Servius was generally aware of the chronological relationships among the various literary figures, the distinction drawn today between archaic and classical usage was not functional in his work.^[39] Further, a necessary corollary derives from this repeated testing of Vergil's language for the obsolete: as in the identification of figurae , the identification of antique dicta has a prescriptive purpose. The basic relationship between the function of figurae and the function of archaism in the commentary can be stated fairly simply: as the demarcation of figurae is an attempt to deal with deviant usage synchronically, by applying the standards of correct usage to the author as though he were a contemporary, so the identification of archaism is an attempt to isolate such deviations diachronically, by constructing a temporal barrier between the author and the student.

Since the two approaches have such similar goals, we should expect them to be expressed in similar language. In fact, the associations that

[38] Corruption of natura : see the note at 2.195, p. 177. Change of the nature of parts of speech: cf. the comment at 5.603, n. 26 above. On the mutability of nature, see Barwick, Remmius 184 n. 1; Siebenborn, Sprachrichtigkeit 108-9; Blank, Ancient Philosophy 41ff.; and contrast Varro ap. Diom. 439.17f., above, n. 20.

[39] Lebek, Verba 18 n. 22, has well remarked the importance of understanding that "archaic" in a grammarian's comment means archaic relative to his own time and perception, and likewise the imprudence of accepting such distinctions at face value.

form in Servius's mind as the two approaches melt one into the other can be seen in the trend of his own rhetoric of instruction. In the examples listed below, we can watch Servius's thought pass from the synchronic to the diachronic, with an intermediary blending of the two, from figurative use versus what "he should have said," through figurative use versus what "we now [ought to] say," to archaism versus what "we now [ought to] say." The examples also further illustrate the interchangeability of debuit dicere and the prescriptive indicative dicimus :

figura. . . . nam debuit dicere. . . . (1.16)

figurate dixit. . . . nam dicimus. . . . (6.435)

figura. . . . nam modo dicimus. . . .^[40] (11.73)

antique dictum. . . . nam nunc dicimus, nec iungimus. . . . (6.544)

debuit dicere. . . . ideo ausus est . . . quia . . . apud maiores . . . dicebatur. (9.467)

archaismos. . . . debuit enim dicere. . . . (10.807)

The instability of the distinction is demonstrated by the progression of the notes and is especially evident in the last two. On Aeneid 9.467, as we saw above, debuit dicere and ausus est look to the present and are intended to have their impact on the student; but the explanation, quia . . . dicebatur , looks to the past: it effectively isolates Vergil's usage and at the same time negates any suggestion that he was in reality bold; his usage, a function of his being one of the maiores , appears bold only when measured against the current state of the language. In the comment on Aeneid 10.807, dum pluit in terris, ut possint sole reducto / exercere diem , the operation is even more striking:

DVM PLVIT hic distinguendum: nam si iunxeris "dum pluit in terris," erit archaismos: debuit enim dicere "in terras." tamen sciendum est hemistichum hoc Lucretii [6.630] esse, quod ita ut invenit Vergilius transtulit.

Here the text is in effect moved into the present and punctuated as though it were a contemporary work in order to arrive at what should have been said and avoid an archaism.^[41] The blending of the two ap-

[40] Cf. 1.75, notanda . . . figura: frequenter enim hac utitur. nam quod nos . . . dicimus, antiqui dicebant. . . , concerning the use of the ablative for the genitive; see further p. 186.

[41] On the series of notes to which 10.807 belongs and with which it must be read to be understood, see further pp. 188f. and n. 55.

proaches is inevitable and derives from the system Servius inherited, in which the categories of auctores (associated with figurative usage) and antiqui (the sources of archaic usage) had long since fallen together: auctoritas and vetustas are for Servius essentially one and the same^[42] and are equally under constraint.

Like the subordination of description to prescription, this fusion of auctoritas and antiquitas , of figures and archaism, necessarily diminishes the precision of Servius's statements and the subtlety or consistency of his response to the text.^[43] The fusion should not, however, obscure Servius's real sense that the antiqui used a language alien, in some fundamental ways, from his own. More than three centuries earlier, Quintilian had observed, "If we compare the language of the ancients with our own, almost everything we say nowadays is a figura " (Inst. 9.3.1). Servius would have agreed, although he would have altered quidquid loquimur to quidquid loquebantur in the second half of the statement.

The strain that shifts in usage produced is perhaps most evident when Servius is faced with a corruption in the received text. In such places we can see him struggling mightily but in vain to heave a line across the abyss: thus, commenting on Aeneid 9.486-87, nec te tua funere mater / produxi (as printed by Mynors, with Bembo's emendation), he attempts the following:

NEC TVA FVNERA MATER id est "funerea": nam apud maiores "funeras" dicebant eas ad quas funus pertinet, ut sororem, matrem. nam praeficae, ut et supra [6.216] diximus, sunt planctus principes, non doloris. "funeras" autem dicebant quasi "funereas," ad quas pertinet funus.

The first sentence here offers a wholly fictive explanation according to what "the ancients used to say"; the second introduces an irrelevancy recalled from an earlier comment; and the third simply restates the first by way of conclusion.

Most often, however, Servius's command of natura and his awareness of its changes provide a more useful (if still shaky) bridge. Consider, for example, his note on the difficult bit of phrasing at Aeneid 11.149-50, feretro Pallante reposto / procubuit super :

[42] The only important exceptions to this statement are the so-called neoterici , i.e., the post-Vergilian poets, especially Lucan, Juvenal, and Statius, who were auctores without being veteres. As exceptions, they caused Servius some difficulty; see Kaster, "Servius." The early collapse of auctoritas and vetustas into a single category was well noted by Barwick, Remmius 215.

[43] For treatment Of the same usage now synchronically, now diachronically, compare Servius's remarks at 3.359 with those at 12.519; similarly 8.168 (on the use of bina ) in conjunction with the notes at 1.93 and 313.

FERETRO PALLANTE REPOST.O posito Pallantis feretro: nam antiptosis est.

Servius explains the phrase by invoking antiptosis, the use of one case in place of another (here, ablative for genitive). The explanation is evidently wide of the mark, but the reason for Servius's error is sometimes misunderstood. Perhaps naturally, the modern reader assumes that the technical term is used "as a joker card" to avoid the problem, and that "the ablative for Servius has the meaning of the possessive genitive."^[44] Servius's own thoughts, however, move in precisely the opposite direction: Servius is certain of the nature and function of the ablative and genitive in his own language, and is also certain that in the language of Vergil and the antiqui these amounted to something very different. It is Vergil for whom the ablative had the meaning of the genitive, as Servius had occasion to remark early on:

1.75 PVLCHRA PROLE. . . . notanda tamen figura: frequenter enim hac utitur. ham quod nos per genitivum singularem dicimus, antiqui per septimum dicebant, ut hoc loco "parentem pulchra prole," id est, "pulchrae pro]is."

The belief is hardly unique to Servius and could only be reinforced by Vergil's repeated practice.^[45]

That the manipulations of Vergil could appear more odd at a distance of four hundred years than they do at a distance of two thousand—quite as odd as the archaic usage (in our sense) of Ennius—is a quirk of language and history not always fully appreciated. Thus Richard Bentley, observing Servius's note on Aeneid 10.710, PASTVS pro "pastum." nam supra ait "quem": ergo antiptosis est , reacted with characteristic vigor to what he took to be a grammarian's sleight of hand: "What the hell is that antiptosis?"^[46] Both Servius and Bentley were attempting to treat a passage where, again, the received text was corrupt:

[44] Williams, "Servius" 52.

[45] With Servius's interpretation of 11.149, cf. the alternative explanation of 2.554-55, hic exitus illum / sorte tulit , noted by DServ., quidam "exitus sorte" pro "sortis" tradunt, ablativum pro genetivo. Such interpretations were no doubt encouraged by the grammarians' belief that the Romans invented the ablative so that it could share the burden of the original, i.e., Greek, genitive; cf. Pomp. 171.18-20. For the recognition of archaism implied in claims of antiptosis, see especially Nonius Marcellus, Book 9, "De numeris et casibus," which is wholly concerned with instances of antiptosis and presents vividly the distance between the language of the antiqui and the perceptions of late antiquity.

[46] Quae malum illa antiptosis! in his comment at Hot. Epod. 5.28.

aper, multos Vesulus quem pinifer annos defendit multosque palus Laurentia silva pastus harundinea.
(Aen. 10.708-10)

True to himself and to his sense of independence and authority, Bentley emended pastus to pascit or pavit (the former is printed by Mynors). Servius, also true to himself, interpreted the passage according to his own sense of the language and its changes, relying on an inference drawn from passages where Vergil does use the nominative in place of the accusative.^[47] Bentley here was right, Servius wrong.^[48] But we should understand that in such cases Servius is using the technical terms not to conceal his difficulties but to acknowledge the discontinuity between the Latin of the antiqui and his own. The technical term is simply an economical device provided the grammarian by his profession. Its meaning is condensed, its function in the commentary both expressive and effective: it simultaneously reveals to and impresses upon the student the distance that separates him from Vergil. Offering a guarantee that carries the weight of Servius's institutional authority, the technical term both conveys and enforces the lesson to be learned.

The examples in the last three paragraphs are extreme cases, finding Servius at or near the point of helplessness, and show his method at its worst, measured against modern expectations. But the extreme cases only highlight the normal practice. Servius's narrow historical perspective and his largely prescriptive concerns anchor him in the present moment, the nunc , of his teaching. His purpose is to anchor the student in the same rather strange slice of time. So one finds early in the commentary the following note, of a very common type transparent in its intentions:

1.3 MVLTVM ILLE. . . . "ille" hoc loco abundat. est enim interposita particula propter metri necessitatem, ut stet versus: ham si detrahas "ille," stat sensus. . . . est autem archaismos.

This comment, with, for example, that on Aeneid 5.540, PRIMVM ANTE OMNES unum vacat , or on 5.833, PRINCEPS ANTE OMNES unum vacat ,

[47] See DServ. on 2.377, delapsus ; cf. also the Grammatica Vergiliana attributed to Asper (Thilo and Hagen, eds., Servii . . . Commentarii 3.534) on 1.314, obvia. Cf. also Servius at 11.775, cassida , for the accusative used in place of the nominative; morphologically his description of Vergil's neologism is not far off.

[48] Note, however, that Servius was only more obviously wrong than the large majority of modern editors, who solve the problem by punctuating after Laurentia , thus producing for pastus a variety of colon whose disposition in the hexameter is thoroughly at odds with Vergil's practice, as Bentley himself was later to show in his comment on Lucan 1.231; see also Townend's valuable paper "Some Problems" esp. 339-43 for Vergil.

should be understood as aimed at the tendency of the common language to add unnecessary words or to use synonymous pairs for intensification.^[49] The note, with its concluding sentence, is meant to suggest, "This sort of excess baggage [abundat ] is obsolete: that is not the way we [ought to] speak or write nowadays." And it is with the message of this note in mind that one must read, as Servius's students would have heard, the long series of notes of the abundat or vacat type that follows.^[50]

The purpose and net effect of such notes is to place the unwanted usage of the auctores firmly in the past:

1.176 RAPVITQVE IN FOMITE FLAMMAM paene soloecophanes est. nam cum mutationem verbum significat, ablativo usus est. sed hoc solvit aut antiqua circa communes praepositiones licentia, ut est [Georg. 1.442] "conditus in nubem," contra [Aen. 2.401] "et nota conduntur in alvo" . . . ; aut "rapuit" "raptim fecit" flammam in fomite, id est, celeriter.

The note regards the coordination of prepositions with the case system and is cast in effect in the form of a quaestio —Why is this not a solecism?—to which two solutions are offered. The second is specific, explaining that the ablative is correct by current standards because no change of place occurs.^[51] The first is general and more interesting, the invocation of the antiqua licentia : the standard that today would mark the usage as a solecism does not apply to the antiqui , whose language did not draw the same distinctions Servius's does in the use of communes praepositiones.^[52] The note, which has as a variant the type found on Aeneid 9.467,^[53] is meant to fix the distinction in the minds of Servius's students even as it exempts the antiqui by drawing a line between the past and the present.

[49] For the text used to exemplify and so to undermine bad habits of common speech, see n. 27 and below, p. 189. On pleonastic intensification, see Löfstedt, Late Latin 21-24.

[50] Compare 1.12, TYRII TENVERE COLONI deest "quam," with the generalization concerning what amant antiqui dicere vs. nos exprimimus : the generalization is clearly meant to enforce the proper use of the relative pronoun; cf. the large number of other notes of the deest type, the complement of the vacat type.

[51] Cf. Servius at 2.401 (cited in his comment), and at 5.332, 6.51, or 10.305.

[52] The lesson is expanded at 1.295 to include the assertion that the natura of certain communes praepositiones has changed over time. For the antiqua licentia , see also 1.253, on the phrase in sceptra reponis , 6.203, on super arbore ; and Serv. 419.27-36, 443.7-10, Pomp. 275.19ff., 276.28ff. Cf. Chap. 4 at n. 101, on Pomp. 237.10-22, contrasting the confusiones antiquitatis (the product of a time when adhuc indefinita erat ista ratio ) with the regula of Donatus.

[53] Debuit dicere. . . . ideo ausus est . . . quia . . . apud maiores . . . dicebatur ; see pp. 180, 184 above.

Yet the blurring of distinctions in this matter—and above all the haphazard use of in with the ablative and accusative—was much more characteristic of late Latin than of the ancients.^[54] Once more the undesirable practice of common speech is put off on the antiqui. And to provide reinforcement, a series of scholia proceeds from this note, reminding the student that the lack of proper distinction belongs to the past, is obsolete, archaismos.^[55] We have come a long way since then, the grammarian says: the movement of the language under the grammarian's guiding hand, toward greater refinement and regularity and away from ancient confusion or licence or harshness, is not doubted, and is a source of no little satisfaction.^[56]

Servius's insistent and complacent didacticism makes his observations unreliable and sometimes bizarre, but not disingenuous. The distinction needs to be emphasized not only for a fair reading of Servius but, more important, for the reasons underlying it. When a usage is explained as arising metri necessitate , when Servius suggests what Vergil debuit dicere , when he comments on what "we say," when he distinguishes "what we say now" from archaism, the text of Vergil and the general state of the language are subordinated to Servius's sense of his own function and authority. Instead of being real objects that one tries to explain or describe historically, text and language become ciphers, assigned whatever validity or significance Servius chooses. The choice is complex and subjective, but it is not a matter of raw and conscious manipulation: it is expressed impersonally, through appeals to natura and the use of technical terms, the guarantees provided by his institutional niche; but Servius not only accepts those guarantees, and the authority they provide, as useful tools, he absorbs them into his personality. Servius believes what he says—about Vergil, the antiqui , the language, and nos —because he simply cannot believe otherwise. He has been fused with the institution he represents.

That the impersonal guarantees, rules, and authority have all been internalized is evident when we find the workings of Servius's mind displayed unself-consciously in that habit of projection we have already seen at work in Pompeius.^[57] Like Pompeius, Servius most reveals himself when he explains someone else's actions and motives. To bring the

[54] Cf. Adams, Text 54-55.

[55] See 2.541, aut archaismos aut. . . , the general and specific, similar to 1.176; 6.639, archaismos , contradicting Donatus; 9.442, with additional reference to 9.347; 10.387; 10.838; and 10.807 (archaismos vs. debuit dicere ), the matter of punctuation discussed above, p. 184.

[56] Beyond the references in nn. 50-53 and 55, see, e.g., Diom. GL 1.374.5ff., 400.1ff., 427.15, 435.22ff. (and cf. 371.16ff., 381.12ff.); Mar. Vict. Ars 4.84 Mariotti (p. 85).

[57] See Chap. 4 pp. 166ff.

chapter to a close, we can examine three passages that both betray this habit of mind and draw together several of the points remarked above—the use of the text as dummy, the nuance of the prescriptive indicative, and Servius's conception of his own status.

The first example is straightforward:

8.435 TVRBATAE pro "turbantis": nam timuit homoeoteleuton et fecit supinam significationem.

The note combines exegesis (it explains and justifies Vergil's use of the wrong participle) with prescription, and is thus a variation on the metri causa or debuit dicere type; whereas the latter sort is essentially negative, obliquely warning the student against a given usage, the comment on Aeneid 8.435 is largely positive. Servius projects his own values and concerns onto Vergil in order to inculcate the lesson in his students: as Servius is and as he would have his students be, so Vergil "was afraid of homoeoteleuton" (the collocation turbantisPalladis ), because homoeoteleuton represents a vitiosa etocutio , a flawed form of expression, to be avoided in polished speech or writing.^[58]

This instance of projection does not require much comment, but it should be compared with our second example, where the same tendency is present in a more interesting if less obvious form. The scholium involves the normative force of dicimus. The person who serves this time as the medium of Servius's message is Valerius Probus:

10.444 AEQVORE IVSSO pro "ipsi iussi." et est usurpatum participium: nam "iubeor" non dicimus unde potest venire "iussus." sic ergo hic participium usurpavit, ut Horatius verbum, dicens [Epist. 1.5.21] "haec ego procurare et idoneus imperor et non invitus." ergo satis licenter dictum est, adeo ut huic loco Probus [hic corruptum] alogum adposuerit.^[59]

[58] For criticism of homoeoteleuton as a vitiosa elocutio , see the notes at 4.504, 9.49 and 606; cf. Pomp. 304.10, antiquum est hoc totum, hodie nemo facit. siqui fecerit, ridetur. With the function of the phrase timuit homoeoteleuton at 8.435, cf. 3.663, 10.571, 11.464, 12.5 and 781, and Schindel, Figurenlehren 27. With timuit , cf. also Serv. 409.17f.; and Pompeius on the fear of Donatus, Chap. 4 pp. 167f.

[59] The text given is that of Thilo, who seems to have made the best of the general corruption in his manuscripts by treating the phrase hic corruptum (appearing only in M of the manuscripts he used) as an interpolated note originally intended to describe the state of Servius's text itself, where the nonsensical reading .a. longam in M (a longam or ad longam in the other manuscripts) occurs instead of alogum , an emendation of Burmann. Alternatively, hic corruptum could have found its way into the text at an earlier stage as a gloss on an original reading alogum.

The didactic intent of the note concerns the form iubeor and could be paraphrased, "We do not [ought not] use the passive iubeor or forms derived from it. Look: Vergil did, and his use is so odd that Probus even marked the passage as flawed."^[60] But Probus likely did nothing of the sort. This is not to say that Servius invented Probus's annotation, but that the concerns of the two men were probably not so congruent as Servius in the urgency of making his point came to suggest. Where Servius reacts to the question of morphology, Probus was probably reacting to the sense, the figure of thought—the epithet transferred in using aequore iusso in place of socii . . . iussi.^[61]

Several considerations suggest that Servius has gone astray in referring to Probus. First, the thought rather than the verbal form seems to have attracted earlier comment. That is the concern of the (no doubt traditional) gloss that begins Servius's own note; moreover, this particular figure of thought seems to have stood prominently in collections of such passages: when Macrobius's Servius enthusiastically recites expressions allegedly coined by Vergil, it occurs near the head of the list (Sat. 6.6.3), where again it is the transfer of the epithet that is noted. Second, in contrast to the foregoing, there is the singularity of Servius's own teaching. His condemnation of iubeor (and so iussus ) is unique among the grammarians, but his citation of Horace's imperor suggests his train of thought clearly enough.^[62]Iubeor is proscribed according to the principle that a verb governing the dative in the active voice (e.g., impero tibi, invideo tibi, obicio tibi; cf. Diom. GL 1.399.13-32 for a full account)

[60] For the alogus , see Isid. Orig. 1.20.17: alogus nota quae ad mendas adhibetur ; the alogus is listed among the critical signs, but is not glossed, in the Anecdoton Parisinum (GL 7.533-36). On this and the other notae said to have been used by Probus, see now Jocelyn, "Annotations" (I-III). My thanks to Prof. Jocelyn for showing me the proofs of "Annotations III," with his discussion of Aen. 10.444 (p. 472), in advance of publication: though he doubts (per litt. ) that we can establish anything at all about Probus's reasoning here, there is no great difference between us in the interpretation of Servius's remarks.

[61] So already Ribbeck, Prolegomena 151; but his prior assumptions regarding Servius's learning, derived from the portrait in Macrobius, made him unwilling to believe that Servius himself could have been concerned with the mere grammatical point. He therefore regarded the note as an interpolation; a similar conclusion was reached, for slightly different reasons, by Georgii, Antike Äneiskritik 454-55. Cf. also Scivoletto, "Filologia" 117.

should be construed impersonally in the passive: imperatur mihi , not imperor. Since iubeo came to be used with the dative under the influence of impero , it should be governed, Servius reasons, by the same rule: iubetur mihi , not iubeor —a bit strict, but certainly unexceptionable Latin. Servius's prohibition of the passive participle iussus is a further, less-than-thoughtful regularization.

But that leads to the Final consideration. The attempt at regularity that inspired Servius's remarks on Aeneid 10.444 is unlikely to have appealed to Probus.^[63] In line with his taste for older authors unfashionable in his day, Probus's views ran in the direction of anomaly. His opinion concerning "those rotten rules and cesspools of grammar" is on record and accords with Suetonius's portrait of the man as something of an anomaly in the world of the grammatici , with interests and practices that deviated from the norm.^[64] It would seem certain that the licence that disturbed Probus concerned the idea, the nonsensical () collocation of aequor and iussum (the bidden plain).^[65] But Servius seized upon the grammatical form; finding in one of his sources a reference (probably vague) to Probus's annotation, he instinctively assumed their concerns were identical and saw support for his own eccentric position on the question of what "we say." Servius's treatment of his scholarly predecessor is precisely the same as his treatment of Vergil.

Servius's capacity for misunderstanding or misrepresenting his sources has been remarked before,^[66] although not for the reason involved here. The note on Aeneid 10.444 takes us beyond casual manifestations of carelessness or animus to a distortion that, like the nuance of dicimus , is built-in and automatic. Conditioned by Servius's devotion to his professional role, the distortion is virtually a reflex, and as such brings us close to the center of Servius's identity.

[63] Although Aistermann, De M. Valerio 11, connected the scholium at 10.444 (= frg. 36 = frg. 98) with the view attributed to a Probus in cod. Paris. lat. 7491 fol. 92 (= GL 4, xxiii-xxiv = frg. 97), the latter probably does not go back to Valerius Probus but represents an inference drawn from Probus Inst. art., GL 4.156.33-157.3, on verba neutralia.

[64] See Chap. 2 at nn. 84, 85.

[66] See Goold, "Servius" 134-40, concerned mostly with pure blunders; and cf. Zetzel, Latin Textual Criticism 105f., suggesting that Servius's notes at 3.535 and 636 are a malicious distortion of Donatus.

We can perhaps take the last step by looking at our third example, another instance of projection, which seems to define Servius's view of his own status:

4.9 INSOMNIA TERRENT et "terret" et "terrent" legitur. sed si "terret" legerimus, "insomnia" erit vigilia: hoc enim maiores inter vigilias et ea quae videmus in somnis interesse voluerunt, ut "insomnia" generis feminini numeri singularis vigiliam significaret, "insomnia" vero generis neutri numeri pluralis ea quae per somnum videmus. . . . sciendum igitur quia, si "terret" dixerimus, antiqua erit elocutio: "insomnia" enim, licet et Pacuvius et Ennius frequenter dixerit, Plinius tamen exclusit et de usu removit.

Servius's note on the variant readings is set squarely amid a minor bog of Latin lexicography, the distinction between the feminine singular insomnia , "sleeplessness," and the neuter plural insomnia , "(disturbing) dreams."^[67] The lexical point, however, is not the central problem here, but the final clauses of Servius's note, sciendum . . . removit. These must ultimately derive from the elder Pliny's Dubii sermonis libri VIII and are included by Servius to inform his students that the feminine singular insomnia would involve an archaic form of expression.^[68] The precise moment when the usage became obsolete is pinpointed, in Servius's understanding, by the magisterial act of Pliny—exclusit et de usu removit.

The statement and the idea behind it are intriguing: why—and more to the point, how—did Pliny treat the word so that it was excluded and removed from use? How did he express himself? We cannot know for certain, and there is room for various conjectures concerning the distinction Pliny made.^[69] It does seem most likely, however, that Pliny's differentiation of the two ambiguous forms, one feminine singular only, the other neuter plural only, was intended primarily to emphasize the

[67] See Getty, "Insomnia" (similarly DeRuyt, "Note"); and the lengthy rejoinder in Ussani, Insomnia esp. 77-113. The disposition in TLL s.vv. insomnia (-ae , 1935.75-1936.61) and insomnium (-i , 1937.70-1938.76) is the most sensible both in classifying the forms and in noting uncertainties.

[68] Della Casa printed insomniam , the reading of cod. F. of DServ., in her edition of the Dub. serm. (frg. 15, "insomniam" enim . . . removit ) and may have been correct: Servius normally accommodates the case of the word he quotes to the syntax of his sentence. Regardless of whether one reads insomnia or insomniam in Servius, however, the context makes it plain that he means the feminine singular.

[69] For example, Barwick, Remmius 206, suggested that the statement represented a preference for ratio , analogy, over vetustas. Yet the principle of analogy scarcely seems relevant to this problem: more significant are the examples Barwick adduced (p. 207) to show Pliny's special interest in consuetudo.

distinction in meaning, which is the center of attention in Servius and other grammarians as well. In that case, the distinction was probably grounded in Pliny's sense of consuetudo —the usage current in his own day—set against vetustas.

Pliny, in other words, was probably attempting to do no more than clarify an existing situation. In the literary language, the feminine singular was an archaism well before Pliny's day; it is attested only in the older republican poets, of the second and early first century B.C. , and thereafter in the archaizing authors of the second century A.D.^[70] The neuter plural though, appears to have been used regularly in the literary language of the first century A.D. , including that of Pliny himself, and to have enjoyed even greater currency in ordinary speech.^[71] If it is reasonable, then, to believe that Pliny's remarks simply recognized and defined the status quo , we might even suggest how Servius found those remarks transmitted in one of his sources—probably something along the lines of the following: Plinius [or sic Plinius or Plinius ait ]:^[72] "insomnia," licet et Pacuvius et Ennius dixerit, penitus tamen de usu recessit [or exclusa est or remota est ].^[73]

The precise form of the notice is not crucial; in distinguishing the usage of the antiqui , Pliny no doubt used some such phrase as hodie non utimur or abolevit or in usu non est , the kind of phrase that abounds in Servius. The point is this: the magisterial act—exclusit et de usu removit —was probably not Pliny's at all, but the product of Servius's interpretation, the act of a Pliny created by Servius in his own image, with his own prescriptive use of such phrases as hodie non utimur in mind. The chain of events suggested above, it is fair to say, accurately reflects both Servius's method and his self-image. There is no question who, in Servius's mind, has the final say in the life of the language: the simple observation of another man concerned with the language is translated by Servius, removed in time and imbued with the sense of his own authority, into an act of verbal extinction. Ipse dixit.

The grammarian's control of the language was something very personal. He was, to be sure, following a professional tradition of long standing when he offered his students a version of "Received Standard

[70] For insomnia (-ae ), see TLL 1935.75-1936.61.

[71] See Getty, "Insomnia" 21-22; TLL s.v. insomnium , 1939.9-10.

[72] The formula "Plinius: . . ." is of the kind commonly used to introduce the views of an individual in, e.g., the Scholia Veronensia.

[73] Cf. Servius at, e.g., 7.626; 9.4; or 12.298, TORREM erit nominativus "hic torris," et ita nunc dicimus: nam illud Ennii et Pacuvii penitus de usu recessit ; and Pomp. 187.10-12, legimus in Capro. . . . etiam Naevius, Attius, Pacuvius, omnes isti utuntur hoc exemplo. tamen ista de usu remota sunt. The citation of Caper is particularly suggestive, since Caper in his turn used Pliny as a major source.

Imperial" Latin,^[74] expressed in the impersonal terms of his craft—natura, regula , and the like. But as he filtered that version through his own idiosyncratic preferences, choices, and distinctions, the grammarian presented and thought of himself as the maker of the lingua aetatis suae , superior to the claims of auctoritas or antiquitas. Those dissatisfied with the grammarian's personal control could circumvent it only by insisting upon a higher authority: that of God, for example,^[75] or the more diffuse authority of the maiores. Macrobius's Servius took the latter course in the incident from the Saturnalia with which we began this chapter; and for his efforts, the grammarian there was roundly abused by the youth Avienus for purveying the obsolete.

I have already emphasized the radical difference in this regard between the figure Macrobius created and the man who speaks in the commentary. It remains to underline one additional point. When Avienus demands that the participants in the symposium use praesentia verba , the aetatis suae verba (Sat. 1.5.1-2), he is demanding in effect that they speak natural, regular Latin, the Latin covered by the nunc dicimus of the commentary.^[76] In other words, despite the conflict between youth and teacher that Macrobius imagined, Avienus speaks much more in the manner we should expect of a student of the real Servius than of an opponent; were Avienus not meant to prove himself a basically decent sort, it would be easy to imagine him behaving like grammarians' pupils who delight in pointing out what Vergil debuit dicere.^[77] Avienus's rudeness in his dash with Servius is part of his characterization, a prelude to the broader education he receives in the symposium.^[78] But that Avienus should speak as he does at the outset is appropriate in another respect: his initial

[74] The phrase is Löfstedt's, Late Latin 48.

[75] See Gregory the Great's challenge in Ep. 5.53a: situs motusque et praepositionum casus servare contemno, quia indignum vehementer existimo, ut verba caelestis oraculi restringam sub regulis Donati. Cf. Smaragdus of St. Michel (s.IX): "I disagree with Donatus, because I hold the authority of Scripture to be greater," cited by Robins, Ancient and Medieval Grammatical Theory 71 (his translation).

[76] In addition to the remarks above, see Pomp. 186.34-187.16 for usus falling together with ars ; cf. the revealing comment of Porphyrio on Hor. AP 70-71: "cadentque / quae nunc sunt in honore vocabula, si volet usus": hoc est, ratio loquendi. [usus ] nihil enim aliud est quam regula sermonis Latini. Petschenig correctly secluded usus in the note; the whole scholium is clearly a comment on usus in Horace's verse. See also ps.-Acro ad loc. , and Brink's valuable note, Horace 2.158-59.

[77] Thus the pueri scorned by St. Augustine, C. Faust. 22.25 (PL 42.417), in a comparison with those who find fault with the prophets: similes sunt, qui in magnis ista reprehendunt, pueris inperitis in schola, qui cure pro magno didicerint nomini numeri singularis verbum numeri singularis esse reddendum, reprehendunt Latinae linguae doctissimum auctorem, qui dixit [1.212] "pars in frusta secant." debuit enim, inquiunt, dicere "secat." On debuit dicere as a criticism, see n. 31.

[78] See Kaster, "Macrobius" 242ff.

deficiencies are precisely comparable to those of the plebeia grammaticorum cohors , whose inadequate knowledge and narrowly defined expertise Macrobius repeatedly criticized. In the Saturnalia , the members of that cohors are despised for shutting themselves off, as though in a box sealed by their ignorance of the culture's roots, whereas the idealized grammarian uses the language to bring past and present together. But the Servius of the commentary limits the language's scope and personally guards all approaches to it. He is in fact just another member of the troop, using the box—his institutional niche—as his position of strength.

In these last two chapters we have reviewed some of the elements that contributed to this position of strength: the accumulation of learning preserved in the tradition on which the grammarian could rely; the confidence in the rational ordering of the language's nature and in the greater sophistication, relative to the ancients, that it brought; the ability to apply one's own learning and ratio to decide between competing views, or even to add a new or more solid piece to the great edifice here or there; and the anxious need to protect the nature of the language—and, closely linked to it, one's own expertise—from assault. There was in all this a nice cooperation between the grammarian and his tradition. The tradition fortified the grammarian in the authority and security of his niche; the grammarian preserved the tradition and paid it the compliment of his improvements.

Such cooperation made for an enduring equilibrium. The grammarian was not about to criticize the tradition in any basic and general way or to be encouraged by his fellows to do so. The obvious urge to be right was independent of any drive to say what was both true and fundamentally new in the conception of the language or in the methods of discussing it. In its broad outlines and in much of its detail, the truth had already been found, in Diomedes' phrase, through "the brilliance of human talent" (GL 1.299.3). Such confidence perhaps led to what modern scholars often see as stagnation and a failure to evolve. But in the eyes of the grammarian, that stagnation was nothing other than the stability of lasting achievement; the failure to evolve, a satisfaction with what was already effective. It is worth remembering that even the most significant innovation in the late-antique ars —Priscian's treatment of syntax—is self-consciously presented as an infusion of earlier learning from a branch of the tradition that his Latin colleagues had previously neglected:^[79] what had worked before would continue to work, but would work even better for the adjustment.

[79] Prisc. GL 2.2.2ff., on the incorporation of the work done by Apollonius Dyscolus and Herodian.

From this point of view, grammar's failure to evolve is not attributable to some failure of nerve or intelligence but is a measure of its success. It remained as it was not because it was exhausted but because it worked so well and so smoothly. Perhaps the grammarians' satisfaction with the forms of analysis and the conceptual categories they had inherited over the centuries, and likewise their confidence in the familiar ordering of the language's nature, would have been shaken if they had had to confront more unruly data, derived, say, from the vulgar language of the market or of the suburban countryside. At very least, their thinking might have been modified and might have been forced to move in new directions. But of course those strata of the language not only received no sustained and systematic attention; they were effectively beneath interest. Indeed, they were not just beneath interest; they were what the inherited forms of analysis and conceptual categories, with their heavily normative emphasis, were meant to rise above. And that normative emphasis derived in turn from the embedding of the grammarians' position of strength in a larger structure of status and honor. So, after examining the grammarians' understanding of their skill and of their authority within the confines of their niche, we are reminded that this niche did not exist in a vacuum. In the next chapter we will ask how the grammarians' authority served them when they moved beyond the classroom to make their way in the world at large.