Editing the Commedia: a short history

2010 saw the publication of a digital edition of the Commedia overseen by Prue Shaw. Written in the first quarter of the 14th century, Dante’s masterpiece traces the poet’s journey through the topographies of the afterlife. The poem is divided into three cantiche (Inferno, Purgatorio, Paradiso), each composed of 33 cantos, themselves written in hendecasyllabic tercets.

The Commedia has a long and controversial editorial history. This is hardly surprising given the work’s resonance throughout centuries and thus the number of witnesses of the poem, over 600 including complete manuscripts and early printed editions. Scholars have deployed a range of editorial strategies in an attempt to settle on a critical text of the Commedia. In this section, I will summarise these strategies in brief, focusing on those elements of each that remain relevant to the edition under review.

The length of the text and the number of witnesses has made it difficult to compare versions of the text in their entirety. At the end of the 19th century, Michele Barbi sought something of a compromise: he selected a series of lines – the so-called 400 loci (Bartoli, D'Ancona, and Del Lungo 1891; Barbi 1891) – drawn from various points in the Commedia. Scholars were invited to check these lines in those manuscripts of the poem to which they each had access and to send a record of them to the Società Dantesca Italiana. Unfortunately, only a handful of these scholars replied and Barbi’s initiative turned over very little material.

Following the Vandelli (Vandelli 1921) and Casella (Casella 1923) editions, Giuseppe Petrocchi drafted a new Edizione Nazionale in 1965 on the occasion of the seventh centenary of Dante's birth (Petrocchi 1965). Petrocchi chose to limit his project to manuscripts that dated before 1355; the project’s theoretical underpinning was that witnesses drafted after that date were likely contaminated, largely because of the scribal and editorial endeavours of Giovanni Boccaccio, who copied out the entire Commedia at least three times, introducing variants from other copies. Petrocchi's critical text is an attempt to reconstruct the so-called antica vulgata in use during the three decades following Dante's death rather than an archetype; this last concept is hardly applicable to a text which was almost certainly released in sections.

Antonio Lanza's 1995 edition had a different aim: following Bedier's prescripts, the editor reproduced a bon manuscrit – copied in Florence in 1337 –, referring to early Florentine manuscripts in order to make amendments only where strictly necessary (Lanza 1995).

In 2001, Federico Sanguineti published a new edition using a small pool of witnesses: seven manuscripts known as ‘the Sanguineti seven’. His stemma consists of two branches: α and β; to simplify, manuscripts from Tuscany form the α branch, while those from Northern Italy form the β branch. As a general rule in textual and linguistic transmission, the peripheral geographical areas tend to be more conservative than the centre; Florence had early on become an active centre for reproductions of the Commedia, while northern manuscripts tended to be less innovative. Sanguineti decided to use the only second branch witness – ms. Urbinate Latino 366 of the Vatican Library (Urb), written in Emilia Romagna – as the base-text for his critical edition. It is worth clarifying that ms. Urbinate’s solitary position in the β branch is due to the limited overall number of witnesses Sanguineti chose to mobilise and to his choice to move ms. Rb (now disjointed in two voumes: Inferno, Purgatorio at the Biblioteca Riccardiana, while Paradiso at Biblioteca Nazionale Braidense) from the β to the α branch. Critics drawn to Sanguineti's edition have repeatedly expressed their concern regarding the scarcity of readings offered as a proof of validity of his stemmatic hypothesis and their distribution along small sections of the poem (e.g., Segre 2002).

This last step in the editorial history of the Commedia is particularly relevant to the edition under review. Sanguineti was originally part of the Commedia project, whose goal was, in part, to test the validity of his stemma. At the time of the project’s inception, the scholars involved anticipated a positive result for Sanguineti. In Shaw's words, 'it seems important to emphasize that the conclusions we reached are not those we expected to reach when we started out on the digital Commedia project [...]. The result turned out to be more interesting and complex'.

Introduction: Shaw's edition

The Commedia Project officially ran from 1998 to 2010, with funding from The British Academy, The Arts and Humanities Research Council, The Modern Humanities Research Association and the Rockefeller Foundation. The venture’s main actors were Prue Shaw (editor), Jennifer Marshall (research assistant) and Peter Robinson (responsible for the information technology side of the project). As mentioned in the foreword to the edition, the genesis of the project dates back to the early 1990s; the original idea spread between Europe and Australia, and the project was finally structured between 1998 and 2001, when Federico Sanguineti's edition appeared. In 2010, Shaw’s Digital Edition was released online and on DVD. The present review deals only with the online edition – however, there is no difference in content between the two versions.

Shaw's digital edition has much to recommend it: it provides critics with access to seven of the most important witnesses of the poem in both extremely accurate transcription and high-definition images; it questions the soundness of Sanguineti’s stemma by putting it to the test of phylogenetic analysis and de-stabilizes assumptions surrounding the validity of the position of ms. Rb – the basis of Sanguineti's critical text; it provides word collation and tools for investigating variants distribution across the witnesses. Finally, it presents high quality editorial materials on textual and technological aspects.

The decision to limit the scope of the project to the Sanguineti seven manuscripts alone has invited some criticism (Inglese 2012). However, the editors were interested in Sanguineti's hypothesis from the offset precisely because ‘such a small number of manuscripts would make a computer project a feasible possibility’ (see below). The selection reflects a paradox well known to textual scholars dealing with large manuscript transmission and with long texts: though computers do indeed allow human beings to process great quantities of data quickly, preparing materials to be computer-readable might be time consuming to the degree of nullifying the advantages gleaned from the former. Every project must of necessity reach a compromise with its ambition (for a different strategy in applying digital technologies to the edition of the Commedia, see Renello 2013).

Witnesses description, transcription and encoding

High quality images are available for all manuscript leaves, except for the ms. Urb, due to copyright restrictions.

The witnesses’ description is very accurate, examining codicological, palaeographical and linguistic features, enriched by images illustrating the discussed phenomena. Ms. Transcription notes on individual witnesses provide further information specific to each of them.

Image/Text. Witness menu.

Transcription criteria are well documented in the General transcription note. For instance, it includes protocols for recording different kinds of deletion, catchwords, interpolated lines or blank spaces. The transcription is diplomatic (the text is copied without normalization, corrections or emendations), it is not graphetic (different forms of the letters are not recorded) and abbreviations are expanded in italics. The only exceptions to the transciption’s fidelity lie in the manuscript’s punctuation, which is not recorded, and in the absence of a distinction between i and j. The seven manuscripts and two editions are fully transcribed.

Layers menu (Ash).

Different layers of transcription can be shown, highlighting the interventions of different hands, as in the case of the Ashburnham manuscript (Florence, Biblioteca Medicea Laurenziana, Ms. Ashburnham 828).

Layers menu (Mart).

The Braidense witness (Milano, Biblioteca Nazionale Braidense, Aldina AP XVI 25) is an edition published by Aldo Manuzio in 1515, and annotated thirty years later by Luca Martini with variants from an early manuscript no longer in existence. The Aldine text and Martini’s additions can be seen either together or separately.

The witnesses are fully transcribed and encoded. It is worth mentioning another major web resource for scholars interested in reading Dante in manuscripts: the Dante online portal of the Società Dantesca Italiana. The only manuscript fully transcribed in both projects is the Trivulziano (Milan, Biblioteca dell'Archivio storico civico e Trivulziana, Ms. Trivulziano 1080).The double check may be useful, at least for spelling details: e.g., Inf, II, 2: DO toglieua vs SHAW togleua; and Shaw is more consistent in the preservation of the u/v instability. Nevertheless, considering the time needed to transcribe the whole poem, it is worth discussing whether it is worth having two projects independently carry out this task.

Barbara Bordalejo has authored the introductory chapter on the encoding system. The edition provides numerous examples of the encoding of scribal deletion, problematic readings, glosses and alternative readings, substitution of one reading for another, aspects of layout, etc. The texts have been encoded in Collate-style markup (specific to the collation tool Collate, see below), and later transformed into a customized version of XML-TEI. Starting from the transcription guidelines elaborated by the Società Dantesca Italiana for the Dante Online website, the project created a model with which to separate 'the text of the document' and 'the variant states of the text'. Bordalejo observes that the main goal of this new transcription system is to present a clear distinction between the text of the document (i.e. what goes in the lit tag: the exact series of marks upon the page) and how the editor (or the transcriber) interprets the different stages of development of the text (i.e. our understanding of the text as originally written and then altered). Bordalejo, Commedia, Appendices: The Encoding System

Here is an example:

<app> <rdg type="orig">dura</rdg> <rdg type="c1">duro</rdg> <rdg type="lit">dur<hi rend="underdot">a</hi>o</rdg></app>

The first and second reading tags (<rdg>) mark the variant states of the text: the original and the corrected. The third <rdg> registers the text of the document. TEI elements have not here been used as suggested by the TEI Guidelines and little use is made of the TEI Representation of Primary Sources module.

Word collation. Metrical analysis.

Another level of markup regards the metrical system. David Robey's independent work on the encoding has been incorporated and is available with a click on each line. It is the editor's intention that this should help the reader understand the effect different variants might have on the scansion of the line. The marked phenomena are accented syllables, synalephas, the opposite dialephas and diaereses (for an extended discussion of Dante's poetic practice, see Robey 2000).

Collation and phylogenetic analysis

Word Collation. Original spelling forms.

Collating texts of witnesses drawn from different regions and periods inevitably yields variations in spelling. Shaw chose to delete these variations for the collation, standardizing the texts’ spelling in accordance with Petrocchi's critical text – since Sanguineti's own critical text is based on a manuscript containing linguistic forms specific to Northern Italy. Problems concerning spelling, segmentation, rhyme and editorial practices are discussed in detail in the Word Collation chapter. Original spelling forms can be seen immediately by selecting the corresponding button in the Word Collation page.

Image/Text. Apparatus.

The collation process is run with Collate, a program developed by Peter Robinson in the late 1980s (Robinson 1989). The complete collation results can be seen in the Apparatus, available from any word of any manuscript.

Word collation.

Collation results are also displayed in the Word Collation section, line by line and using Petrocchi's edition as the base-text.

The collation generates a single XML-encoded file containing the collation result for the entire Commedia. Data is then stored in standard NEXUS file format for processing with PAUP (Phylogenetic Analysis Using Parsimony). The program uses the record of agreements and disagreements among the witnesses ('taxa', in evolutionary biology) at precise sites of variation ('characters'), to generate a phylogram: an unrooted tree. The length of the branches corresponds to the degree of divergence between witnesses. PAUP, which has proven to be particularly suited to the purposes of textual critics, uses sophisticated methods to find the most 'parsimonious' evolutionary tree: considering all possible bifurcating trees, it identifies the one that requires the smallest number of changes (Windram et al. 2008; Roos, Teemu, and Heikkila 2009; Young 2014). It is worth remembering that evolutionary trees are unrooted, i.e. they have no orientation. For the purposes of textual criticism, this means that the computer fails to provide answers at the highest levels of the stemma.

Variant map.

For every line, a Variant Map is available, showing the result of the phylogenetic analysis.

The data is processed multiple times: once for the entire poem, again for each cantica and yet again for sections within each cantica. The phylogenetic analysis shows that there is no divergence in 88% of the Commedia text. It also produces trees for Inferno and Purgatorio that are 'indistinguishable from those for the whole text, while the tree for Paradiso shows a slight variation' (Introduction). On the basis of their experience with the materials and the analytical method, the editors state: however, one could not assert that part-publication [...] of these cantiche of the Commedia [...] did not happen at all. Simply, the phylogenetic analysis suggests no evidence for it. Shaw, Commedia, Introduction

A comprehensive account of the conclusions drawn from phylogenetic analyses is beyond the scope of this review; following an exhaustive account of the editorial history and textual problems of the poem, philological results are presented on the Introduction page (and particularly in the Manuscripts and computers chapter and the following chapters) and on the Phylogenetic analysis page. In Shaw's words, the project’s most significant finding is the clarification of the position of Rb: Rb is shown unequivocally to be a collaterale of Urb, and not a member of α as Sanguineti maintains. This inevitably has important knock-on effects for the restitutio textus. Shaw, Commedia, Introduction

In his review, Giorgio Inglese adds nuance to this conclusion, suggesting that 'the appearance of a peculiar relation [between the manuscripts Urb and Rb] is just a consequence of the selection of the witnesses' (Inglese 2012). A reading which seems specific to those two manuscripts can in fact appear in others which are not taken into account in the edition. Inglese highlights the value of the electronic collation that allows for an in-depth evaluation analysis by providing readers with the chance to check the editor's conclusions. He follows to observe that, as far as it regards the restitutio textus, Shaw’s hypothesis (Rb has a β text) leads to the same results than his own (Rb has an α text, contaminated with β): Rb and Urb agreement represents β.

Judging variants. Editors vs machines?

A frequent question in Digital Humanities is whether digital technologies have altered methods of scholarship and, if so, how. For instance, how does one reckon with textual variants in the field of scholarly edition? The question inevitably arises wherever scholars seek to track the transmission of a work and to produce a critical text. In the so-called Lachmannian tradition, the editor is required to distinguish between variants and errors and to consider only the latter in drawing relations between witnesses. The trees produced in textual criticism using evolutionary biology methods, however, draw upon all the evidence – that is variants which may or may not include errors. For instance, in her analysis of the relationship between the ms. Urb and the ms. Rb, Shaw presents the following evidence: errors in common (counted as errors by Petrocchi but accepted by Sanguineti) and errors separativi of α (which isolates Urb and Rb), which amount to ten 'monogenetic variants linking Rb and Urb'. In addition, the editor considers small errors and variants according to the following useful taxonomy (for a similar list of types of polygenetic variants, see Brandoli 2007): substitution of a word, change in word order, addition or omission of the definite article, addition or omission of the first person pronoun io and/or the third person pronoun el, addition or omission of a small word, singular for plural or vice versa, variants with an extra or missing syllable which affects scansion, resulting in a metrical error or not, different forms of the verb, retouching of short phrases. As Shaw points out, some of these categories are not especially significant in themselves. Singly, they mean next to nothing. But it is the presence of a long series of them uniformly right across the text in a very small number of manuscripts which is striking (and this is surely what Petrocchi’s phrase ‘foltezza di statistica’ refers to at least in part) [...] If the results of the computer analysis are accepted as valid for mss. AHMT and L, there is no reason that they should not be accepted as valid for mss. R and U. Shaw, Commedia, Introduction

A coherent agreement or divergence between witnesses alongside the text, including errors and variants, is evidence that has to be taken into account. Shaw's assertions are less persuasive, however, where she argues that this ability of phylogenetics to create hypotheses of relationships which do not require any prior judgments as to originality is one of its greatest strengths for textual scholars. In classical stemmatics, as formulated by Paul Maas, analysis must be based on shared error alone. Therefore, one must determine at each point which reading is “original", which is “error", before analysis can begin. As well as the difficulty of determining the “original" reading, there is the argument elegantly expressed by Talbot Donaldson: if one can determine the original reading at every point, then why bother with any further analysis? Shaw, Commedia, Phylogenetic Analysis

However, determining which reading is an error in traditional stemmatics does not amount to identifying the original reading. For instance, an editor can establish that a reading that breaks the metrical scheme is an error and still have other witnesses with different, and potentially correct, readings; or a lectio difficilior might be 'trivialized' by all scribes, in the so-called ‘diffraction in absentia’, such that it remains extremely difficult to 'guess' which reading is the original and correct one. Thus what Shaw considers a matter of interpretation, avoidable by recourse to 'new methodology', in fact remains the domain of scholar's expertise.

Fortunately, Shaw seems to contradict her previous statement, declaring that 'assessing the significance of variants is a large part of the editorial process'. Doing it with digital technologies might be a promising field of investigation, which, to my knowledge, has only been pursued beyond the stage of theoretical assertions by few scholars (Barabucci, Di Iorio and Vitali 2014; Cadioli Forthcoming; a work in progress as part of the DiXiT Marie-Curie Network).