Open Book Publishers logo Open Access logo
  • button
  • button
  • button
GO TO...
book cover

11. Beyond Variants: Some Digital Desiderata for the Critical Apparatus of Ancient Greek and Latin Texts

Cynthia Damon

© Cynthia Damon, CC BY 4.0


Texts from the ancient world reach us via a long, complicated process of transmission from copy to copy. As printed today they are at best near approximations of what an ancient author wrote. A critical edition, which presents the text along with the surviving evidence of the transmission process and an editor’s interpretation of it, allows the reader to go beyond a generalised expectation of error and to see whether any given bit of text is secure, or corrupt, or disputed, or weakly supported by the manuscripts that preserve it. No classical text can be read responsibly without one. Yet existing digital libraries of classical texts routinely strip out ‘the surviving evidence of the transmission process and an editor’s interpretation of it’, presenting only the text.1 They do so by the simple expedient of omitting the critical apparatus.

If we are going to reinstate the critical apparatus—as we must if digital editions of classical texts are to serve the needs of scholarship and if digital libraries are to become the go-to repositories of classical texts—we need to understand what the apparatus is. I am troubled by what I see as a trivialisation of the apparatus in the Text Encoding Initiative’s Guidelines for Electronic Text Encoding and Interchange, which are meant to foster standardisation in digital editions. The TEI Guidelines define the apparatus as a repository of variants, and assert that ‘individual readings are the crucial elements in any critical apparatus of variants’.2 For classical texts, at least, a proper critical apparatus is far more than a repository of textual variants: it is a repository of everything that an editor judges necessary for a reader to understand why the text being read is what it is. More precisely, the apparatus is a set of notes designed to foster in the reader an awareness of the historical and editorial processes that resulted in the text he or she is reading and to give the reader what he or she needs to evaluate the editor’s decisions.3

Some of the apparatus content is of course the variant readings in the manuscript tradition. But these variants only yield a text through the operation of an editor’s theory about how the manuscripts that contain them are related to one another and to the authorial original. So the lists of variants have to be understood as embodiments of a theory. In other words, when the editor reports in the apparatus that manuscripts AB have the reading in the text, while CDE and FG have variants, what he or she communicates is likely to be something like this: ‘Given my theory about how manuscripts ABCDEFG are related to one another, the reading of AB has manuscript authority and makes acceptable sense and is therefore printed, whereas the readings of CDE and FG are scribal innovations in the manuscript from which each group is separately descended and are therefore not used in the constitution of the text’. And beneath that message is a theory that defines what ‘a reading with manuscript authority’ is: basically, a reading that may have reached us through a continuous sequence of accurate copies of what the author wrote back in antiquity and may therefore be authentic and (by definition) right.4 This is much too much to write in small print at the bottom of the page for every lemma, but some variation on that reasoning is often implicit in notes that list manuscript variants. Furthermore, for most classical works the manuscript variants do not suffice for constituting the text. At best they allow one to reconstruct an archetype. But archetypes usually postdate the authorial original by centuries, often many centuries.5 As copy succeeded copy corruptions must have entered the text, even if we can no longer trace the process in any detail. The situation is even worse when the process of transmission that produced the extant manuscripts cannot be represented genealogically, which means that one cannot reconstruct an archetype or make strong assertions about manuscript authority. In such traditions the manuscript variants will be even further removed, temporally and culturally, from the authorial text. They have to be evaluated on their merits, and all too often their merits are not enough to generate an acceptable text. In cases like these editors have recourse to emendation or declare the text irreparably corrupt. All of this is presented in the critical apparatus, which is therefore a repository not of variants but of arguments (in the best sense of the word) about variants.6 We need to find a way to embody the arguments as well as the variants in the digital critical apparatus.

Arguments in the apparatus

I begin by illustrating something of the range of types of arguments found in the critical apparatus of a classical text with examples from recent major critical editions. Which types will appear in any given apparatus and their relative prevalence will depend on ‘house style’, editorial policy, transmission history and the nature of the work being edited, but in designing a digital apparatus we should be prepared to accommodate all of them and more.7

The simplest sort of note reports variants whose distribution enables the editor to reconstruct the archetype and arrive at a printable text. In a tradition with three branches, for example, such as that of VegetiusEpitoma rei militaris, the agreement of two against the third will give you the reading of the archetype:

3.9.3 perscribam εβ : de- δ8

Here ε, β and δ represent the heads of the three principal families of the tradition. The editor, following the agreement of ε and β, prints perscribam. The variant describam in δ is an innovation. This note offers both an explanation of the reading in the text and evidence relevant to the assessment of the editor’s stemma. (As will become clear as we proceed, the spatial compass of apparatus notes is usually reduced by the omission of things that can be assumed, such as the repetition of -scribam here; this increases the demand on the reader considerably.9 Fonts, too, are part of the code: Roman font is used for words belonging to the text, italic font for words or symbols contributed by the editor.)

If a reconstructable archetype has an unacceptable reading, the editor may want to explain why it has been discarded and/or what has taken its place. Thus, again from Vegetius:

1.4.1 imbibuntur δ : imbuu- εβ, vix latine (cf. TLL 427.56–60)

The archetype was the source of imbuuntur in ε and β, but that reading yields dubious Latin that is only paralleled by other doubtful passages listed in the reference work cited in the note, the TLL.10 The reading in δ, imbibuntur, which the editor prints, is an innovation of some sort, but acceptable Latin. The editor of Vegetius does not justify his preferred reading here except by showing the argument against imbuuntur.

But you will also find notes with explicit justifications, such as this one in the newish OCT of ApuleiusMetamorphoses:

4.2.3 proprio F, corr. AU11

This tells the reader that the word in the text of 4.2.3 most closely resembling proprio, namely proripio, is misspelled in the manuscript F, which is the archetype of this tradition, but spelled correctly in two descendants of F. The correction was presumably made independently by A and U, since it does not appear in other descendants of their common parent from a generation between themselves and F. The editor’s ‘corr.’, meaning correxerunt, explains the genesis of the reading in the text. However, it may also provoke the reader to query the editor’s explanation. If, for example, one feels that proripio is not a plausible correction for proprio, especially not one that would have been made twice, one may start looking for another explanation.

A very common type of note that requires considerable mental effort from the reader arises when each branch of a two-branch tradition preserves a construable reading from the archetype. The two branches in theory have equal authority, so how does the editor choose? Consider the following note from the OCT of Columella’s Res rustica: 7 relinquendique SA : retin- R12

Here SA and R represent the two branches. Both relinquendique and retinendique suit the context, which is about the honourable acquisition of wealth. The question is, once the wealth has been acquired, is the next step passing it on to your heirs, which would be the point of relinquendique, or holding onto it for yourself, via retinendique? Here the editor prints the reading that is less likely to have arisen from the other (or some common original) as an innovation, relinquendique. No explanation is given, or needed, so long as it can be taken for granted that the reader understands the principle of the lectio difficilior.13 Retinendique simplifies the phrase since it does not require one to infer an indirect object, as relinquendique does.

Similarly challenging is the related scenario in which one branch of a two-branch tradition preserves a correct reading from the archetype, while the other goes astray but one or more of its constituents retrieves the correct reading. We can see this in a note pertaining to MacrobiusSaturnalia, where the two families are headed by α and β:

3.10.6 docte] te β (recte V)14

The word before the square bracket, docte, is what the editor prints. The note states explicitly that β reads te, and, armed with the knowledge—if we have read the edition’s preface—that the tradition has two branches, we infer that docte is the reading transmitted by α. In β this has been corrupted to te, which makes no sense in the text. But ‘recte V’, meaning ‘V has it right’, tells us that the scribe of V, a descendant of β, has recovered docte, either by conjecture or by contact with the family of α. These two explanations have very different implications for the transmission history of this text, so the note is a piece of the evidence necessary for evaluating the editor’s account of that transmission.

A different sort of challenge for the reader arises when a variant that needs reporting comes from an external source. Thus the editor of the Saturnalia informs us that Macrobius and Aulus Gellius quote Caesar’s famous dictum about avoiding unusual words ‘as one would a reef’ using different terms for ‘unusual’:

1.5.2 infrequens] inauditum Gell.

This note tells us that the archetype of Macrobian tradition transmits infrequens, but also that the reading might not be right.

Some notes contain arguments about matters of interpretation. This is obviously a vastly extensible category and is for the most part excluded from the apparatus. But not always, and for this we should be grateful.

A note that efficiently explains the syntax of a difficult passage can save much head-scratching, as in this note from the OCT of Columella: 17 ut enim [...] uisos] oratio obliqua ἀνακολούθως (cf. Varr. Rust. maiores [...] praeponebant [...] ut ruri enim [...] desidiosiores putabant)

By identifying the construction as an incompletely expressed indirect statement (‘oratio obliqua ἀνακολούθως’) and by supplying a parallel passage in which one can see in putabant the sort of governing verb that seems implicit here, the editor allows us to read on unperturbed—unless we feel that it would be better after all to supply putabant or something like it in the text.

Another situation where interpretation may be felt to deserve mention in the apparatus is where ancient scholarship bears on the text. For example, on a description in Virgil’s Aeneid that is presented in the text as part of a character’s speech, the editor notes that the ancient grammarian Servius knew of a competing interpretation that took it as an utterance by the narrator:

6.573 “tunc [...] portae [...] alii hoc a poeta dictum volunt [...] alii continuant narrationem” Serv.15

A string of words running from tunc to portae, says Servius, is attributed by some authorities to the poet, by others to the character in the narrative; the editor follows the second group. This note does include a variant—Servius read tunc where the manuscripts used for constituting the text offer tum—but the variant is not the point.16 The editor is prompting the reader to evaluate his editorial decision to treat the text as character-speech. This kind of challenge is present whenever the editor includes interpretative aids, be they ancient or modern.

The claim on the reader’s attention is even more pressing when the editor indicates that the archetype’s text is acceptable but nevertheless suspect for one reason or another. An editor has a variety of signals from which to choose. One can simply convey doubt. Thus in a passage from Ovid’s Metamorphoses where the transmitted reading is unexceptionable in form and syntax but at odds with its context the editor draws attention to the passage with this brief note:

8.262 mitis habebatur suspectum17

The reader will find mitis habebatur in the text and infer that it is transmitted by the manuscripts, but he or she is invited by the editor to view it with suspicion.

Where possible and appropriate the suspicious editor may suggest an alternative, either one found in a manuscript or a modern scholar’s proposal. This is the point of a note like the following from the Metamorphoses, which tells us that one or more manuscripts from the thirteenth century, here represented by the siglum χ, read faciat where the manuscripts used to constitute the text read capiat:

9.749 capiat] faciat (cf. 1.469) χ, fort. recte18

From this brief annotation one understands that the editor went looking for a way to improve upon the text’s capiat and found it in a place that he would not normally cite. He cites it here on the strength of a parallel passage earlier in the poem (1.469) and indicates with his ‘fort. recte’, meaning ‘perhaps rightly’, that the innovation in χ may retrieve what Ovid originally wrote.

A word can also be flagged as suspect by a list of attempts to emend it, even if the editor accepts none of them. Thus the puzzling word reluctabant in the description of a ragged and filthy robber in ApuleiusMetamorphoses gets the following note in the OCT edition:

7.5.3 reluctabant F, def. Armini1 (notione relucendi), aliter Helm (‘i.q. specie discrepabant) aliter Hijmans3 (‘were having a wrestling match’) : relucebant uel relucitabant Gruterus

The editor prints the reading of F, but the fact that three scholars explain it differently (aliter) while a fourth proposes two different emendations for it (uel) gives the reader plenty of reason to distrust the text here.

The editor may also want to float a suggestion of his or her own. Thus in Vegetius, where the archetype and the text read duritia, the reader is alerted to its metrical oddity by the editor’s proposal to substitute a synonym:

1.6.4 an duritate ob numeros?

Duritate, the note explains, yields a more regular prose rhythm.

Or the editor may suggest that the text would be better off without the offending bit. That would be the case if, say, it made its way into the transmitted text from a source such as a marginal comment. Thus the note on Met. 9.749 that we looked at a moment ago in fact begins with the following warning:

9.749 del. Heinsius

In addition to the line’s textual problems, of which capiat is just one, the line is a banal and self-contained aphorism, so it was excised (del.) by a great Renaissance editor of Ovid, Nicolaas Heinsius. Readers would do well, the editor indicates, to be cautious about embracing it as echt Ovid.

I could go on to illustrate notes that explain substantive repairs to the transmitted text, or notes that indicate by a lacuna or a crux that the editor has despaired of repairing the text, or notes about punctuation or orthography or illegible codices or the host of other issues that editors address in the critical apparatus, but the examples above suffice, I think, to show that the apparatus for a classical text is much more than a repository of variants. Its notes constitute a highly evolved form of philological argument.

Decoding the apparatus

The form may in fact be too highly evolved to move easily into a new medium. Apparatus notes cannot simply be read; they have to be decoded, not only by the expansion of abbreviations and the filling of omissions, but often also with the help of concepts and theories that are presented elsewhere in the edition or in the scholarly literature. In traditions where the relationships among manuscripts are essentially genealogical, for example, the stemma that charts those relationships is a key to the meaning of many apparatus notes.

The interpretative consequences that arise when a textual string in an apparatus has been properly decoded with the help of a stemma can be seen from the following example, taken from Loyen’s 1970 Budé edition of the letters of Sidonius Apollinaris, a fifth-century bishop from Gaul.19 The passage comes from Sidonius’ travelogue description of Ravenna, and I give Loyen’s translation:

Ep. 1.5.5 insuper oppidum duplex pars interluit Padi, cetera pars alluit.

“En outre la cité elle-même est coupée en deux par un bras du Pô, tandis que le reste de ses eaux la baigne”.

cetera F : certa cett. codd.

Ravenna is a two-part town, Sidonius tells us (oppidum duplex), because one branch of the Po runs through it (interluit), splitting it in two. He then adds something more about the river, which not only ‘flows through’ (interluit) but also ‘flows alongside’ (alluit). The apparatus reports the manuscript variants as follows: a manuscript called F reads cetera, the rest read certa. But what does it mean when Loyen prints F’s cetera in his text? (Earlier editors either printed certa or marked the passage as irreparably corrupt.20) It is not a simple choice between manuscript readings. A stemma for this text is given in Figure 11.1 below. (As is traditional in stemmata the capital letters represent actual manuscripts, the Greek letters represent lost manuscripts whose existence is inferred from the presence of shared errors in their descendants.)

Fig. 11.1 Stemma for Sidonius, Epistles 1.5.5.

The earliest recoverable phase of the text is represented by α, the archetype, which passed down its readings to L and β, which passed its readings to γ and δ, and so on. The underlying principle is that the readings in, say, δ, are essentially those of α, together with the innovations introduced in the copying process. The innovations become visible when independent copies of a single model differ from one another. The editor’s job is, in the first place, to identify the innovations at each stage of transmission and to work back to the reading of the archetype. What this stemma tells you is that the readings that reach F from α have been successfully transmitted through β, δ and ζ. For our passage it would have to work like this: α had cetera, β copied it correctly and L made an error, writing certa. Then δ copied cetera correctly and γ, the source of T and M, made an error, writing certa. Then ζ copied it correctly and P made an error, writing certa. And finally F copied it correctly, and C made an error, writing certa. The hypothesis that four scribes independently made the very same copying error is extremely improbable. Therefore the reading in F is probably the innovation. That is, although it is in the manuscripts, it has no manuscript authority. So when Loyen prints it he is saying that the archetype was corrupt and that this scribal innovation is the best available repair. The whole argument is implicit in the simple apparatus entry ‘cetera F : certa cett. codd.’; but you have to decode it properly.

It is hard work to elicit arguments from the brief and cryptic notes that constitute a typical critical apparatus. But without the apparatus the text is a deceptively smooth and satisfying surface. We need both the variants and the arguments that make them interpretable. If a digital apparatus can provide something more user-friendly than the traditional print form, so much the better for the future of textual scholarship.

The need is in my view urgent. At present someone who wants a critical apparatus can go to the physical volume—provided that one exists with a competent apparatus criticus and provided that it is accessible. But a generation hence, when the use of online texts will have become the norm, readers will not expect or want to go to the physical volume. And a digital library that has equipped its texts with some kind of apparatus will be the only source that provides them with a fully adequate text. Back in 2000 Michael Reeve, a distinguished textual critic, observed that ‘Until the apparatus can be restored, there is a danger that electronic texts will be trusted further than any text merits even if accurately reproduced’.21 What readers need is a digital edition that not only gives them access to a given string of words but also enables them to understand why the string comprises just those words in just that order.22 The balance of this paper offers some thoughts on how to achieve such a thing.

Where to start?

In a provocative opinion piece posted in 2012 Paolo Monella asked why classicists have been slow to embark on the project of creating digital critical editions for the works of Greco-Roman antiquity.23 The answer that he proposed seems to me largely on-target, namely that the value that classicists place on the documentary basis of our texts is limited and pragmatic rather than ecumenical. That is, if I want to make a literary or historical argument about what Caesar wrote, any innovations in the text of Caesar by a medieval scribe are going to be a distraction if not a snare. So I rely on editors to give me the soundest possible text of Caesar’s writings, as well as the evidence for it. And my interest in that evidence is pragmatic: does it support the reading that my argument is based on or not? If I try to imagine making a digital critical edition on the model widely considered to be fundamental24—where you start with page images of the witnesses, proceed to transcription and coding, then apply tools to produce collations and to link images and versions, and then you add annotations and so on—my first feeling is despair. I simply cannot imagine that any classicist—or funding agency—will invest the kind of time and money that would be needed to create digital critical editions of the whole corpus of Greek and Latin literature.

My next feeling, however, is exasperation. ‘And why would we need so vast an investment?’ After all, with classical texts, at least, we are far beyond the pioneering textual labours implied by this model. That work has been done and is recorded in the critical apparatuses of generations’ worth of editions. Those apparatuses and editions are by no means perfect, but to redo the whole process that produced them would be a massive waste of effort. And yet—and yet. Classical texts risk being left behind at this juncture precisely because they started out ahead. What classicists need, I would argue, is a way to give digital form now to the mature state of textual scholarship represented by print editions, while leaving open the possibility of adding the underlying image and transcription data when and if opportunities arise. In other words, we need to start in the middle, not at the beginning. Can the TEI critical apparatus module help us do so? I have to say that it is sobering to report that one of the most tech-savvy classicists alive today, Donald Mastronarde, chose not to use TEI encoding for the apparatus of his Euripides Scholia project, which is otherwise TEI-based. As he explains in the discussion of the work’s XML structure:

The apparatus criticus is an area in which I have decided not to use the TEI mechanisms for apparatus criticus readings and variants, because in a project of this kind it seems to me that it would involve an unjustifiably large overhead of markup. I believe the information familiar to those who know how to read the apparatus criticus of a classical text can be provided in textual segments. This does mean that one will not be able to take my XML document and process it to produce a text that reflects the textual choices and errors of a particular witness, which probably would be possible with a more elaborate markup of readings and witnesses with pointers to specific words in the text. Such a project would require more personnel and a much larger budget, and I don’t think the benefit would be worth the cost.25

Greg Crane’s Open Philology project may give us a starting point.26 He proposes to scan and make available versions of legacy editions that will be OCR-able with varying but ever-improving degrees of accuracy. In my view these images, and more particularly the images of the apparatuses, could serve as a foundational layer for transcription and encoding. That is, rather than start with the images of the manuscripts themselves, seductive as these are, we should start with the cryptic and crabbed text of the critical apparatus, which records the results of editorial collation of the manuscripts and the arguments that give those results meaning. The challenge will be to find new coding (TEI or other) that is flexible enough to enable meaningful collation of highly idiosyncratic expressions of comparable information.

A picture will help me convey what I mean by idiosyncratic. Below I give the apparatus entries from five modern editions on the first line of poem 13 in the first book of Ovid’s Amores (Figure 11.2). In other words, a small subset of the data for one line of 750 or so in Amores I, which is one book of thirty-six by Ovid. A tiny fragment of the data needed for a digital critical edition of Ovid’s work. And look how complicated it is. Almost everything here pertains to the preposition ‘a’, which all five editors print in the text.27

Ovid, Amores 1.13.1: Iam super Oceanum uenit a seniore marito

Fig. 11.2 Apparatus notes on Ovid, Amores 1.13.1.

Basically what all of these apparatus notes say is that the preposition a is omitted by the principal manuscripts and is present in manuscripts of lesser authority (relevant text in green above). The fourth editor goes further, providing pointers to the arguments for printing the preposition (blue). He and two others also summarise or point to arguments against printing a (orange). And finally the very generous fourth editor reports a Renaissance conjecture (purple; the material in black justifies the capitalisation of Oceanum).

The editorial decision to print the preposition has significant interpretative consequences, since if the line includes the preposition, the verb uenit is present tense (short e), if it doesn’t, the verb is uenit (long e), a perfect tense. The present tense verb ‘she is coming’ frames the poem as an address to Dawn as she arrives, a plea to the goddess to delay her arrival so that the lover and his beloved can enjoy a few more minutes of darkness before the day’s round begins. A perfect tense verb in the first line frames the poem as a reproach: Dawn has come and I’m going to tell her just what her arrival means for us mortals. To make a long story short, the annotation already available for this line needs to accompany the line in its digital future. It is also worth noticing that the important information becomes visible when you compare the apparatus entries, not when you compare the texts of the editions, all of which print the same reading.

When I applied the ‘starting from scratch’ approach to this poem in the Juxtacommons program,28 here is what emerged as the apparatus entry for the line:

a] ed.; not in P, S, Y

Because I had transcribed a published text, ‘ed.’, and the three principal manuscripts, P, S and Y, but not the more than 200 manuscripts of lesser authority, I ended up with a report that says that the edition prints a, which is absent from the principal manuscripts. This is significantly less than even the briefest of the existing apparatuses, that of Kenney, which at least tells us that the preposition is present in a class of manuscripts labelled ω. To put it differently, if we are going to spend time and money on digitising Ovid’s Amores for the use of future scholarship, transcribing and encoding the apparatuses would allow us to start in the present rather than back in the humanist era when readers started comparing the medieval witnesses to the works of classical antiquity.

The problems, however, are considerable. The idiosyncratic and coded language of these notes would need to be translated into concepts before they could be properly re-encoded in a standardised TEI or other digital schema.

Consider the first part of the entry in Lenz’s edition:

a<Y(+a Yc)PS

The concepts of omission, addition and correction expressed here are also present in McKeown’s entry:

a : om. PYSCD

But the form in which they are expressed is very different: Lenz uses the mathematical symbols < and + to convey omission and addition, and a suprascript c to convey correction, whereas McKeown uses the Latin abbreviation om., meaning omiserunt, to convey omission and the layout of the note to convey addition, and he uses a lower-case letter (y) to convey correction: little y represents a later hand in big Y. Furthermore, the second entry has more information than the first: it reports the readings of ω and manuscripts labelled C and D as well as those of P, Y and S. But the first entry, when viewed in toto, has information not in the second, bibliographic information, also rather cryptic in form.

a<Y(+a Yc)PS, vgl. Ehw., Burs. 167, 1914, 187; Eisenhut, Gnomon 25, 1953, 447

Lenz sends you for starters to someone referred to as ‘Ehw.’, presumably in a publication referred to as ‘Burs.’, presumably to a volume or fascicle of that publication numbered 167 and published in 1914, and presumably to p. 187 of that volume. Any digital reader of that line would appreciate a translation, and not just because it is in German, unlike most apparatuses.

Doing the conceptual translations needed to correlate the five apparatus entries shown earlier will require a lot of work. However, even a minimal sort of coding that would simply allow you to stack up the entries as I did on the page here would be better than nothing.29 So here is a more precise version of the question I asked before: Does (or could) the TEI critical apparatus module, despite its assertion that ‘Individual readings are the crucial elements in any critical apparatus of variants’, accommodate encoding that prioritises notes over witnesses? I would love to be able to extract readings from the subset of apparatus notes that concern variants, of course, but if I had to choose between recording the readings of five witnesses and recording the notes of five apparatuses I would choose the latter every time.


My aims in the three parts of this brief essay have been (1) to illustrate the inadequacy of the conceptualisation underpinning the current TEI module on the critical apparatus, (2) to demonstrate the power and complexity of the ‘encoding’ already present in the critical apparatus of classical texts, and (3) to encourage a pragmatic approach to the urgent need for digital libraries of critical editions of classical texts. Only with a comprehensive understanding of the content and assumptions of the traditional highly-evolved critical apparatus will we make the right strategic decisions for the future of textual scholarship on the literary legacy of Greece and Rome.30

1 Some openly available digital libraries of Greek and Latin texts: (Greek and Latin); (Greek and Latin); (Greek and Latin); (Latin). Some subscription-based libraries: (Thesaurus Linguae Graecae); (Library of Latin Texts, series A and B).

2 TEI Guidelines for the critical apparatus module are available at, section 12.1.2, [my emphasis].

3 This point is also made in Marina Buzzoni’s chapter in the present volume.

4 Readings without manuscript authority can be right but they cannot be authentic. That is, they can be shown to have entered the tradition through innovation, and may be right if the innovation corrected an earlier error in such a way as to retrieve the original text.

5 For a substantial portion of Petronius’ wonderful Satyrica, for example, a work written in the first century CE, the archetype is a manuscript of the fifteenth century. See M. D. Reeve, ‘Petronius’, in Texts and Transmission: A Survey of the Latin Classics, ed. by L. D. Reynolds (Oxford: Oxford University Press, 1983), pp. 295–300.

6 A point that is well made in general terms by Hans Walter Gabler in ‘Theorizing the Digital Scholarly Edition’, Literature Compass 7 (2010), 43–56,; esp. p. 45: ‘The apparatus entries [...] thus function as argument for the establishment of the edition text’ and on the apparatus as ‘a discourse complementary to that of the edition text’. I have tried to offer some useful specifics.

7 Most of my examples in this first part of the paper are taken from a single series, the Oxford Classical Texts series, to minimise the variations due to house style. And I have chosen them from Latin texts but not Greek ones, to minimise the variations of language. (There is an irreducible minimum of Latin here, since the critical apparatus for classical texts is traditionally presented in Latin, but I have tried to make the point of each example independent of the language of the text and apparatus.) Variations in house style are illustrated in the discussion of the apparatuses for Ovid, Amores 1.13.1.

8 Vegetius, Epitoma rei militaris (Oxford Classical Texts), ed. by M. D. Reeve (Oxford: Oxford University Press, 2004). This late-antique text is unusually well preserved in the manuscript tradition, such that the editor can say that ‘at least one of the three reconstructed witnesses almost always has an acceptable reading’ and ‘it [...] seems unlikely that the tradition derives from an archetype more recent than a copy put into circulation by Vegetius himself’ (p. xlviii).

9 As does Reeve’s assumption that the reader will identify ‘per’ in perscribam as a prepositional prefix and infer that ‘de’ has the same function. That is, he assumes that the reader will supply ‘scribam’ after de, not the sequence of letters that follows the ‘e’ in perscribam. The underlying policy is stated in general terms in the preface: ‘I assume that the reader knows some Latin and expects the scribe [...] to have written a Latin word [...] not a monstrosity’ (p. xlii).

10 The TLL is the Thesaurus linguae latinae, an on-going major Latin dictionary project accessible in print and on-line by subscription at

11 Apulei Metamorphoseon libri XI (Oxford Classical Texts), ed. by M. Zimmerman (Oxford: Oxford University Press, 2012). The manuscript that is the source of all currently extant manuscripts, F, is itself extant but hard to read and damaged in spots, so that its descendants are sometimes called in as witnesses to the text.

12 L. Iuni Moderati Columellae Res Rustica: Incerti auctoris Liber de arboribus (Oxford Classical Texts), ed. by R. H. Rodgers (Oxford: Oxford University Press, 2010).

13 More fully, lectio difficilior lectio potior, a rule-of-thumb asserting that of two readings with equal manuscript authority the more difficult is to be preferred, on the grounds that it is more likely to be authentic, or rather less likely to have arisen from scribal innovation.

14 Macrobii Ambrosii Theodosii Saturnalia (Oxford Classical Texts), ed. by Robert A. Kaster (Oxford: Oxford University Press, 2011).

15 P. Vergili Maronis Opera, ed. by M. Geymonat, 2nd ed. (Rome: Edizioni di storia e letteratura, 2008).

16 To complicate matters further, the lemma for the note (tum demum horrisono stridentes cardine sacrae/panduntur portae) starts in line 573 (tunc/tum) but runs into line 574 (portae) and therefore overlaps with subsequent notes about stridentes and sacrae. The issues arising from multi-word lemmata will be need careful attention when the apparatus assumes digital form.

17 P. Ovidi Nasonis Metamorphoses (Oxford Classical Texts), ed. by Richard J. Tarrant (Oxford: Oxford University Press, 2004).

18 The meaning of the siglum χ is defined in the edition’s introduction (p. xlvi).

19 Sidoine Apollinaire, Poèmes et lettres, ed. by André Loyen, 3 vols. (Paris: Belles Lettres, 1960–1970).

20 Certa is printed in C. Sollius Apollinaris Sidonius, ed. by Paul Mohr (Leipzig: Teubner, 1895); †certa in Gai Solii Apollinaris Sidonii Epistulae et carmina, ed. by Christian Lütjohann (Berlin: Weidmann, 1887). Franz Dolveck is currently reassessing this stemma.

21 M. D. Reeve, ‘Cuius in usum? Recent and Future Editing’, Journal of Roman Studies, 90 (2000), 196–206 (p. 200).

22 I am grateful to Bob Kaster for this neat formulation of the objective of a critical edition.

24 E.g. by Peter Robinson for the Canterbury Tales, ‘A digital edition should be based on full-text transcription of the original texts into electronic form [...]’. Similarly, transcription is step 1 on the workflow proposed by Peter Shillingsburg for the HRIT project:

27 The editions are these: P. Ovidi Nasonis Amores, Medicamina faciei femineae, Ars amatoria, Remedia amoris (Oxford Classical Texts), ed. by Edward J. Kenney (Oxford: Oxford University Press, 1961) (I used this edition instead of its successor in order to highlight the effect of the re-evaluation of Y, which is used for the constitution of the text after 1963); Ovid, Amores: Text, prolegomena, and commentary, ed. by J. C. McKeown, 4 vols. (Liverpool: Francis Cairns, 1987–2012); P. Ovidi Nasonis Amores, ed. by Franco Munari, 2nd ed. (Florence: Nuova Italia, 1955); P. Ovidius Naso, Carmina amatoria: Amores, Medicamina faciei femineae, Ars amatoria, Remedia amoris, ed. by Antonio Ramirez de Verger (Munich: Saur, 2003); Die Liebeselegien, ed. by Friedrich Walter Lenz (Darmstadt: Wissenschaftliche Buchgesellschaft, 1965).

29 As my students and I found when producing just such ‘stacks’ for the Annals of Tacitus in the autumn of 2012, one immediate and unanticipated consequence was the ability to discover and render innocuous the errors and inconsistencies that have accumulated in the legacy of textual scholarship. Another was that the apparatus ‘coding’ became slightly more perspicuous with the juxtaposition of the same information in a variety of formats.

30 I am grateful to my fellow speakers at the 2013 NeDiMAH Experts’ seminar, and especially to the seminar’s organisers, Matthew Driscoll and Elena Pierazzo, for a stimulating discussion of the wide world of digital critical editions, a topic that has been engaging my attention increasingly since 2008, when I organised a panel on ‘Critical editions in the 21st century’ for the American Philological Association. Further stimulus to thought was provided by the ‘Digital Variorum Editions’ project for an NEH ‘Digging into Data’ initiative, and I am grateful to Greg Crane both for the opportunity to be its respondent and for our many discussions on the digital future of classical texts. I have also learnt much from my fellow members of the Planning Committee for the Digital Latin Library, a project co-sponsored by the APA, the Medieval Academy of America and the Renaissance Society of America, especially its director, Sam Huskey, all of whom helped me to formulate desiderata and see opportunities for the ongoing development of critical editions. Especial thanks go to Bob Kaster and Richard Tarrant for many a conversation about the critical apparatus over the years.