1Since the 1950s, editorial theory and practice in the English-speaking world, France, and Germany have increasingly addressed issues of literary genesis: drafts, versions, the writing process, and methods of editorial rendering. While these particular scholarly traditions are quite distinct and will continue to be so for some time to come (see Van Hulle 2004 for a comparative assessment of these traditions), the development of common encoding standards for digital editions supports the formation of a zone of convergence for different traditions. The new chapter 11 of version 2.0.0 of TEI P5 is probably a first step in that direction (TEI Consortium 2011a; cf. the discussion in Brüning, Henzel, and Pravida 2012).
2The new edition of Goethe's Faust consists of four basic parts:
- a digital archive
- a digital genetic edition
- a critically established text of Faust (both digital and print)
- a printed facsimile edition of selected manuscripts1
3The digital archive contains facsimiles, transcriptions, and metadata. On this basis, the genetic edition attempts to reconstruct and visualize the genetic relationships within and between the documents. Diplomatic renderings, as well as presentations of the texts and their variants in different forms of critical apparatus, are to be generated automatically from the basic transcripts. A critical text of Faust will be established by taking into account the complete record of variants. The edition combines two different aims: to meet the standards established by printed scholarly editions and by the tradition of editorial theory, and to adopt recent developments in the field of electronic editions, thereby exploring the evolving possibilities of the digital medium.
4The project's principles of encoding were developed in conjunction with the workgroup (part of the TEI SIG on Manuscripts) that developed the "Encoding Model for Genetic Editions" (Workgroup on Genetic Editions 2010), which formed the basis for the revised chapter 11 of TEI P5. A rethinking of the basic tenets of TEI markup and of our way of adopting the TEI standard for our edition has led us to disentangle one single form of "text encoding" into two different but closely related forms of encoding in order to provide both a diplomatic rendering of the layout of each manuscript (generated from a "documentary" transcript) and a representation of textual genesis (generated from a "textual" transcript). This differentiation complies with the theoretical distinction between record and editorial interpretation as proposed by Hans Zeller (1995): the documentary transcript records inscriptional phenomena, whereas the textual transcript records the genesis of a text.
5Multiple encoding of similar (if not the same) information has some obvious disadvantages. The disadvantages can, however, be dealt with by the use of a sufficiently powerful collation tool. Disentangling two perspectives does not imply that they are independent from one another. On the contrary, their close correlation is of great importance for the reconstruction of genetic relations.
- 2 For a brief account of the history of research on the genesis of Goethe's Faust, see Bohnenkamp et (...)
6Goethe's work on Faust lasted sixty years, from about 1775 until shortly before his death in 1832. Manuscript material comprising about 2,300 pages written by Goethe and his scribes is extant, most of it dating from 1825 or later. Since then considerable effort has been devoted to the investigation of the material. But despite a number of very valuable special studies (the latest being Bohnenkamp 1994) the genetic process still has not been analyzed in its entirety. The only comprehensive account of the textual variants dates back to the nineteenth century (Goethe 1887–88).2 The presentation of such a large body of manuscripts and of a wealth of scholarly knowledge accumulated within a century of assiduous research is a daunting task. The material and the corresponding information should be made easily accessible in a way that enables and fosters further research. Any editorial uncertainties are not to be ignored, and conflicting scholarly opinions have to be considered.
7In the following, we will concentrate on the digital archive and the genetic edition, especially the underlying encoding model, which not only applies the set of tags available in the TEI Guidelines but also extends it at some points.
8The digital archive is a research platform that makes accessible all manuscripts and the authoritative prints of Faust. Facsimiles, diplomatic rendering, and physical descriptions (metadata) of all witnesses will afford the user possibilities of in-depth manuscript research.3 The genetic edition gives an account of the textual alterations within single manuscripts and of the stemmatic relationships obtaining between them. The analysis of both will be visualized using not only various forms of critical apparatus and text presentation but also charts and other graphical representations of data.
9The diplomatic rendering, as well as the presentation of texts and their genesis, is each based on XML-encoded transcripts. Any piece of information to be conveyed to the user has to be put into these transcripts. Therefore, the richness and quality of the information hinge on the constraints imposed by formats and rules of encoding. The editors have adopted the principle that particular problems should be solved according to general and explicit rules of encoding. The encoding principles and strategies devised for the Faust edition should in principle be transferable to editions of other authors with comparable habits of writing. As will be shown, problems that appear to be particular may turn out to be closely related to very general aspects of textuality.
10Throughout this paper, reference will be made to the following example from a revised autograph copy of Faust II (fig. 1), which poses a problem of encoding that may be quite typical for modern literary manuscripts:
Figure 1: Detail from a revised autograph (Goethe and Schiller Archive Weimar, GSA 25/W 1427), Faust II, line 5142.
11The line in the highlighted box of figure 1 reads "Wenn ich es erreichen könnte" ("If I could achieve" [my goal]), written with ink. When revising the manuscript with lead pencil, Goethe modified the text by putting "s" in front of "ich" ("I"), thereby changing the personal pronoun into a reflexive pronoun, struck out the words "es erreichen", and wrote "sie" ("she") and "entschließen" ("to bring [herself] to") above the line.
12The overall conception of an integrated electronic edition requires that all forms of presentation of information on manuscripts and texts be generated automatically from the transcript files. For this purpose, both inscriptional phenomena and textual alterations must be accounted for in detail. Version 1.3.0 of P5 offered a variety of tags and attributes to do this (TEI Consortium 2008, 72–78, 335–74). According to this version of P5, the example given in figure 1 may be encoded like this:
<l>Wenn <add place="above" hand="#g_bl_lat">sie</add>
<del rend="strikethrough">es erreichen</del>
Example 1: Encoding of figure 1 in accordance to version 1.3.0 of P5.
13Example 1 actually reflects the encoding practice of an early stage of the project. The structural divisions of the text and the textual alterations were recorded with the elements shown in example 1: <l>, <del>, <add>, and <subst>. Information on specific material properties—mode of deletion, lineation of the manuscript, scribal hand, writing medium, position of added words—was conveyed by the elements <hi>, <lb/>, and <handShift/>, and by the attributes @hand, @rend, and @place (for the values of the attribute @hand, see section 4.1 below). The transcript contains all information necessary for generating a diplomatic rendering since the approximate spatial position of the written characters can be inferred from it. Likewise, it is possible to deduce the primary and the final wording from the encoding so that some sort of text presentation and critical apparatus may be feasible. However, a closer look reveals that the encoding in example 1 is unsatisfactory in many respects.
14The sequences "sie" and "entschließen" lack any explicit indication of their precise horizontal position. In the relatively simple case of example 1, it might still be possible to generate an adequate diplomatic rendering because the value "above" can be interpreted as "right above." In more complex cases (such as retraced letters or accumulated additions in more than one line or distributed over a line) the result would inevitably fall short of the standards set by printed editions (for example, because the values "margin", "top", and "bottom" for @place are far too vague to sufficiently indicate the spatial position of added words or passages). The set of tags offered in P5 version 1.3.0 is extensive, but it was not designed to fit all the needs of a full-fledged diplomatic edition.
15Despite its lack of precision, the encoding in example 1 is too much focused on single aspects of the writing process to appropriately cover textual alterations. In the encoding, "sie" (she) appears to be merely added, whereas "ich" (I) is, as it were, supplemented by an "s." A thorough account of the textual alterations would have to indicate that one single substitution takes place: a personal pronoun ("ich"), a direct object ("es"), a verb ("erreichen"), and the first person singular conjunctive preterit of können ("könnte"—if I could) are jointly substituted by another pronoun ("sie"), a reflexive pronoun ("sich") together with the corresponding verb "entschließen", and a homograph third person form ("könnte"—if she could). In the encoding of example 1 there is not the faintest indication that all three individual changes made with pencil belong to one and the same textual alteration. The fact that the "ich" is substituted by "sie" is completely blurred by the only superficially adequate "<add>ſ</add>ich". The silent grammatical conversion of the finite verb is not expressed at all, as there is no tagging around the word "könnte" to mark this change.
- 4 At the same time, the "existing chapter on Physical Transcription [i.e. chapter 10] has been exhaus (...)
16In short, the initial way of encoding was neither rich enough to generate a reliable diplomatic rendering, nor did it meet the requirements of a text encoding in the proper sense of the word, i.e., an encoding that respects the linguistic nature of written texts. The need to extend the existing tagset according to genetic research interests has been felt in other editions and projects as well (Saller 2003; Vanhoutte 2002; Pierazzo 2009). To satisfy this need, the Workgroup on Genetic Editions, part of the TEI Manuscripts SIG, drafted an "Encoding Model for Genetic Editions" (Workgroup on Genetic Editions 2010). It has been largely incorporated in chapter 11 of version 2.0.0 and later of P5 (TEI Consortium 2011a).4 It is not our aim to give an overview of all new elements and attributes introduced in the new version of chapter 11. Instead we will discuss below the application of the new tags to the example given in figure 1.
- 5 The term "sediments of the writing process" is adopted from Stetter (1997).
17It is now possible to record the sediments of the writing process5 as well as the textual alterations in a much more differentiated way. Temporal relations or syntactic dependencies of alterations can be expressed properly. As a consequence, the complexity of markup keeps increasing in the process of adapting encoding strategies to editorial demands. From a certain point on, the growth of the encoding complexity seriously impairs the readability of the encoding for humans and makes it hard for machines to process. However, there is still no guarantee that all indispensable information can be adequately stored in the transcript.
18In fact, our project reached an impasse caused by bloated encoding even before the new tags were available. In particular, the encoding was impaired by an increasing number of conflicts, many of them caused by overlapping hierarchies (Renear, Mylonas, and Durand 1996). When it finally proved unavoidable, we decided to split the encoding into two distinct transcripts: a documentary transcript of each page of a manuscript and a textual transcript of the whole manuscript. That is to say, the encoding approach reflected in example 1 was abandoned in favor of a rigorous disentangling (fig. 2).
Figure 2: Encoding perspectives / derivation relationships / different layers of the edition.
- 6 Throughout this text we often gratefully adopt the solutions found for the translation of German ed (...)
19This way out is not simply a reaction to encoding problems. It is closely related to the idea that every document must be considered a material object on the one hand and a medium of textual transmission on the other (cf. Brüning, Henzel, and Pravida 2012). This distinction has deep roots in the scholarly tradition, but the actual use in editorial theory is due to Hans Zeller's authoritative article, "Record and Interpretation," whose German original dates from 1971. It is known to anglophone scholars in its English translation of 1995 (Zeller 1995).6 Zeller distinguishes between the manuscript material encountered by the editor—the "record" (Befund)—and its "interpretation" (Deutung, in a somewhat strict sense of the term; for the sake of clarity, we will speak of "editorial interpretation"). Following Zeller, it is the task of the editor to document manuscripts as objectively as possible and to give an editorial interpretation of them in the form of a well-defined representation of the text and of the genetic process that led to its present form (Zeller 1995, 44). In the German tradition of textual scholarship, observance of this distinction has proven to be the touchstone for editions claiming to attain the highest scholarly standards available.
- 7 This distinction is not to be confused with M. J. Driscoll's "levels of transcription" (Driscoll 20 (...)
20The genetic edition of Goethe's Faust aims to fulfill both objectives. The first aim is to make all of the material relating to the work's genesis accessible by means of digital images. To ensure that the record is not only available but also easily accessible, every image is accompanied by a diplomatic rendering which preserves the original lineation and all other relevant inscriptional features of each manuscript and which can be used as a device for close scrutiny of the facsimile or of the manuscript itself. The diplomatic rendering is based on a documentary transcript (example 2) which describes the spatial arrangement of the inscription, scribal hands and writing medium, manipulations of the written letters (for example different forms of deletions), and special marks such as proofreading symbols and print instructions. The documentary transcript does not allow for any textual information on the structural divisions or textual alterations.7
- 8 D. E. Sattler coined the term "phase" for a temporal segment of the writing process. A segment is d (...)
21The second aim is to provide an editorial interpretation of the record from a distinctly textual perspective, with special emphasis on the development of the text brought about by successive alterations and modifications. Various forms of text presentations and of critical apparatuses are used for the systematic collection, presentation, and analysis of these alterations (example 3). Strictly speaking, the term "textual transcript" is shorthand for "textual and genetic transcript," as the textual transcript serves to satisfy two purposes: to describe the structural subdivisions of the text, scenes, speeches, verse lines, and the like; and to capture the genesis of the text, in particular alterations of the text, bonded variants (in the terminology of Zeller 1995, 49), and also, if identifiable, the phases of the writing process.8 The textual transcript is kept free from any documentary elements and attributes.
22As example 2 shows, the documentary transcript is designed both to record the sediments of the writing process and to determine the layout of the diplomatic rendering shown to the user.
<handShift new="#g_t_lat"/>Wenn<anchor xml:id="Wenn"/>
<handShift new="#g_bl_lat"/><seg rend="inbetween">ſ</seg><handShift new="#g_t_lat"/>ich <mod
hand="#g_bl_lat" rend="strikethrough">es <anchor xml:id="erreichen"/>erreichen</mod>
Example 2: Documentary transcript of figure 1.
23The elements used require some further comment because neither the elements themselves nor the encoding approach in general could be conceived of under version 1.3.0 of P5 (TEI Consortium 2008). Even taking into account the new encoding structures recommended in the revisions to chapter 11 of version 2.0.0 of P5 (TEI Consortium 2011a), the encoding above still may seem controversial in certain respects. Considering the type of the markup used, three different kinds of elements and attributes may be distinguished:
- well-established elements and attributes found in version 1.3.0 of P5 (<handShift/>, <seg>, <anchor/>, @type, @new, @xml:id, @rend, and @hand)
- elements introduced in version 2.0.0 of P5 (<line> and <mod>)
- customized elements and attributes—that is, elements and attributes designed for the specific purposes of the edition (in this case <f:left>)
24The elements and attributes of the first group are commonly used: the values of @new and @hand ("g_bl_lat" and "g_t_lat") combine the values of the three attributes scribe="g" (Goethe), medium="bl" (lead pencil) or medium="t" (ink), and script="lat" (Latin).
- 9 See the XML examples corresponding to figures 14, 16, and 17 in section 11.3.4 of the TEI Guideline (...)
- 10 Conceptually similar phenomena, e.g., interlinear and marginal additions, would have to be treated (...)
25The elements of the second group have only recently been incorporated into the TEI Guidelines and deserve a close look. The new element <line> is analogous to the well-established <p> and <l> but is defined in strictly spatial terms (cf. TEI Consortium 2011a, 1083). Our encoding rules demand that a document be transcribed from top to bottom and from left to right according to the spatial distribution of the written characters on the page. The penciled words "sie" and "entschließen" in figure 1 form an interlinear line of their own. In the old version of chapter 11, the relative spatial position of subsequently inserted characters was indicated solely by means of the attribute @place (<add place="above">sie</add> and <add place="above">entschlieſsen</add>), which proved to be insufficient for more exacting diplomatic purposes. Thanks to the newly introduced <line> element, the manuscript lineation, including interlinear lines, can be encoded directly and explicitly. Instead of a structural element like <l>...</l> or <p>...</p> containing the element <add place="above">...</add>, there are two (or more) separate <line>...</line> elements, one for the original line (in figure 1 "Wenn ich es erreichen könnte"), the other(s) for ulterior additions above or below the line, at the left or right margins, or anywhere else. However, the TEI Guidelines still allow the combination of <line> with @place.9 As a consequence, it becomes possible to inconsistently mix structural and spatial markup and, by implication, to confuse a "document-focused" and a "text-focused" approach. For example, even though an interlinear line can be considered as part of a preceding or following line, additions in the margin or at the top or bottom of the page should not be treated in that way. Adding <line> to the list of well-established transcriptional elements and attributes may therefore lead to inconsistencies.10 Another new element is <mod>. As compared with <line>, the use of the <mod> element supplied with a @rend value of "strikethrough" instead of <del> may seem overly meticulous or indirect. But in cases where any suggestions about the nature and coverage of the alteration are to be avoided, the element <mod> is indispensible. (As explained below, the respective scopes of a positively visible cancellation and a textual alteration do not necessarily coincide.)
- 11 Cf., however, section 11.4 of the TEI Guidelines (TEI Consortium 2011a, 371–74).
- 12 This approach is, in a sense, suggested by the principle of preserving spatial relations with struc (...)
- 13 The documentation of the record, therefore, involves a good deal of editorial interpretation, a fac (...)
26Elements and attributes of the third group (marked with the namespace prefix "f" for Faust) have been introduced where the existing ones did not meet the requirements set by our editorial principles. To encode figure 1, the horizontal position of the interlinear line has to be indicated. The spatial relation to the following line has to be expressed straightforwardly, not indirectly (as, for example, by means of a common coordinate system).11 The spatial distribution of the inscription is not mimicked but represented in an idealized (or, as we prefer to say, structured) way.12 In example 2, the left margin of "entschließen" is left-justified with respect to the left margin of "erreichen", although figure 1 shows a minimal horizontal displacement. The penciled "sie" was clearly intended by the writer to be placed above the inter-word space between "Wenn" and the original "ich"; mere imitation of the appearance of the source document would end up in locating "sie" above the closing part of "Wenn".13 To express the obtaining spatial relations, the left margin both of the inter-word space and of the struck-out "erreichen" is marked with <anchor/>. The attribute @f:left indicates that the left margins of <seg>sie</seg> and of <seg>entschlieſsen</seg> are to be left-justified with respect to the <anchor/> element to which the respective value of @f:left points. The diplomatic rendering is thereby fully determined by the encoding.
27From a textual perspective, however, the very same piece of manuscript inscription looks distinctly different. Here, the characters in ink and in pencil are all to be interpreted as forming one single verse line in which a joint substitution occurs:
<del>ich es erreichen könnte</del>
<add>sie ſich entschlieſsen könnte</add>
Example 3: Textual transcript of figure 1 (simplified).
- 14 It should be kept in mind that the example given is fairly simple. There is no need for the element (...)
- 15 The <l> is provided with a reference to the standard line numbering system (in this case line 5142 (...)
28The textual transcript looks much simpler than the documentary one. This is not because the textual perspective is generally less complex (sometimes quite the opposite is the case).14 Nor is it because two attributes have been left out for the sake of brevity.15 What happens to the passage? A verse line that first read "Wenn ich es erreichen könnte" (if I could achieve [my goal]) now reads "Wenn sie sich entschließen könnte" (if she could bring herself [to do something]). From a textual perspective, as explained above, one personal pronoun, "ich" (I), is substituted by another pronoun, "sie" (she), as the letters s and i, c, h make up the reflexive pronoun "sich" and join the verb "entschließen" to substitute for "es erreichen"; "könnte", i.e., the first person singular conjunctive preterit of können (if I could), is substituted by the homograph third person form "könnte" (if she could). From a documentary perspective, however, "könnte" is as unaffected as "Wenn".
- 16 In the field of digital humanities the word "text" is usually associated closely with the way texts (...)
29To put it more generally: The basic elements of the documentary perspective are written letters, whereas the basic elements of the textual perspective are linguistic units.16 Depending on the perspective chosen, one and the same phenomenon has to be treated in two different ways. The textual transcript, thus, is neither more nor less adequate than the documentary one. They are categorically distinct: the first records the sediments of the writing process, the second records the genesis of the text.
30The outlined encoding approach shares some similarities with the first method of handling non-hierarchical information as outlined in section 20.1, "Multiple Encodings of the Same Information," of the TEI Guidelines (TEI Consortium 2011a, 604–605). This way of proceeding is, apart from its own merits in terms of theory and sound editorial practice, "the simplest method of disentangling two (or more) conflicting hierarchical views of the same information […], each time capturing a single view" (ibid.). Each data file can be processed for the purpose of automatic rendering of different types of visualization, as described above.
- 17 For the reasons mentioned in the previous footnote we prefer the term "marked-up content" instead o (...)
31However, such division of the encoding suffers from the disadvantage "that the method requires the maintenance of multiple copies of identical textual content (an invitation to inconsistency) and that there is no explicit indication that the various views, which might be in separate files, are related to each other: it might prove difficult to combine the views or access information from one view while processing the file that contains the encoding of another" (TEI Consortium 2011a, 605). Before we address the problems mentioned here, we would like to emphasize that the disentangling does not result in "multiple copies of identical textual content;" it leads to only a partial duplication of the marked-up content.17 Whereas the documentary and the textual perspective must be considered as two sides of the same coin, i.e., as two different perspectives on one and the same written text, they cannot be regarded as encodings of the very same information. The textual transcript differs from the documentary transcript not only in terms of markup (as is presupposed in chapter 20 of the TEI Guidelines), but also in terms of the marked-up content: "sie entschließen Wenn ſich es erreichen könnte" versus "Wenn ich es erreichen könnte sie ſich entschließen könnte" (see figs. 4 and 5).
32Yet there is identical content in these two TEI documents and with it the threat of inconsistency. The resulting task is twofold:
- The parts of each literal sequence which should be identical are to be identified (that is, they have to be distinguished from those parts that are consciously designed to vary).
- It has to be determined whether these parts are identical or not (that is, unintended variation must be detected).
- 18 Other variations are less critical. For example, foliations are transcribed only in the documentary (...)
33As in the passage under discussion, variations mainly occur in the context of alterations.18 When there are no alterations, both literal sequences should be plainly identical (this is the case for the initial "Wenn"). Where alterations occur, as in the rest of the line, things are more complex. As far as the deleted wording in the textual transcript is concerned ("ich es erreichen könnte", tagged with the <del> element), the marked-up content of this transcript still has a direct counterpart in the documentary one. This requires isolating the string "ich" in the documentary transcript (instead of "sich"), which is possible if the fact is taken into account that the "s" is tagged as inserted later than the original "ich" (see example 2).
- 19 In this context, "transposition" refers to one of the four kinds of operations needed to transform (...)
34More serious problems arise when the correspondence between interlinear lines and inserted characters (documentary perspective) and additions (textual perspective) is to be detected automatically. In the textual transcript the strings "sie" and "entschließen" (tagged with the <add> element) follow the string "Wenn ich es erreichen könnte", whereas in the documentary transcript they precede it (see example 4). A collation algorithm which handles transpositions would be able to detect the correspondence.19
<line>sie entschlieſsen</line> <line>Wenn <seg>ſ</seg>ich es erreichen könnte</line>
<l>Wenn <subst><del>ich es erreichen könnte</del><add>sie ſich entschlieſsen könnte</add></subst></l>
Example 4: Correspondences between the documentary and the textual transcript.
35An algorithm that fulfills these requirements is presently being developed by Ronald Haentjens Dekker and Gregor Middell (2011). But even if we assume its availability, the textual transcript captures another linguistic level (words, word groups, etc.) beyond the documentary transcript (graphemes). For example, an insertion of a character in the documentary transcript may correspond to the replacement of a whole word in the textual transcript. In our example, the inserted "ſ" in the added "sich" in the textual transcript, which is tagged with the <add> element, lacks a direct counterpart in the documentary transcript: here, the "<seg>ſ</seg>ich" has been segmented into the strings "ſ" and "ich". The correspondent of the documentary "ich" in the textual transcript is "<del>ich …</del>" (see example 3), and if the one-to-one correspondence between the corresponding pairs of strings is preserved, only the "ſ" in the documentary transcript remains as a potential counterpart of "<add>… ſich …</add>" in the textual transcript. The same holds true for the added third person form of können—"könnte"—the occurrence of which does not leave the slightest trace in the documentary transcript.
36At this point the automatic correlation reaches its limits. Manual post-processing of collation results is, as always, necessary. For most parts of the encoding, however, the feared inconsistency can be kept effectively under control by automatic collation. And what is more, the same tool which allows for the reduction of unintended variation will also help "to combine the views or access information from one view while processing the file that contains the encoding of another" (TEI Consortium 2011a, 605).
37When using a scholarly edition, the user will either read the final text and want to look up all earlier versions of a given line and their manuscript source, or will want to follow the genesis of the text or one of its passages from its germ. To attain this aim, all parts of the edition have to be closely connected, and the available connections have to be intuitively and intelligibly visualized (for more details, see Bohnenkamp et al. 2011). Correlating documentary and textual transcripts, as described in this paper, is only the first step towards the general goal of affording the user good ways to navigate between corresponding parts of the digital archive, the genetic edition, and the established text.