University of Virginia Library

A New Approach to the Critical Constitution of Literary Texts
by
Hans Zeller [*]

Independently of one another, in mutual ignorance of the endeavours of the neighbouring discipline, English and German studies have until recently followed quite similar paths regard-the critical constitution of literary texts, though German editors have generally been working without the solid equipment of analytical bibliography. But for about the last ten years a new theory of the establishing of texts has begun to take effect in German editions, and seems to be gaining more and more ground with editors. Seen from the viewpoint of English studies, this new theory might look like a regression from the position held by Greg to the McKerrow phase of editing. Yet the motives behind the emergence and elaboration of the new Germanist theory are not the same as those implicit in such objections as have been raised against the application of the Greg-Bowers principles to the editions of modern American authors. Therefore the recent efforts of Germanists may be of interest to editors of English and American literature, just as the Germanist concept must rely on the application of bibliographical criteria for further expansion and development. It has not by any means attained the level of the applied Greg-Bowers principles, which display a greater refinement and more solid foundations than any other editing procedure known to me.

It goes without saying that the common purpose of both is to elicit, through the distortions of transmission, what the author wrote, and what corresponded to his intention; and there is a common endeavour to proceed according to formulated rules. What distinguishes

more recent German editions both from most earlier editions and from recent English ones is fundamentally a different understanding of the notions of version ("Fassung") and of authorial intention and authority ("Autorisation"), and ultimately a different theory of the literary work and its mutations.

This new understanding resulted from a crisis which occurred when the edition of Goethe's works, planned by the Deutsche Akademie der Wissenschaften in Berlin since 1949, was in preparation. The Academy edition was to have replaced the so-called Weimar or Sophia edition, a historical-critical edition of Goethe's works conceived in the last two decades of the nineteenth century. Just as the decisive phases in the development of English editing have been determined largely by Shakespeare scholarship, so also German editing, with its conceptions and its mistakes, has been shaped by Goethe studies. When Greg's fundamental essay appeared in 1950, Ernst Grumach, who was at the time in charge of the Academy edition, urged the necessity for a new edition of Goethe with a critically established text.[1] His plea was based above all on criticism of the Weimar edition and its canonisation of one particular authority, in this case the final revised edition ("Ausgabe letzter Hand") supervised by Goethe.

Over a period of years (1825-1830), and with Goethe's approval and guidance, his philological assistants, above all Göttling, brought uniformity to the linguistic form (for instance, inflexion) and to the orthography of his works, and revised the punctuation; Goethe himself only occasionally intervened and worked with them himself, thus conferring upon the shape of the work what has been termed merely "passive authorisation", as opposed to the "active authorisation" which alone can be binding for the editor. To the strong unauthoritative textual overlay caused by this process and approved by Goethe, were added those corruptions which the texts trailed with them. These were augmented in the customary manner from edition to edition, often for decades: distortions which could usually have been discovered only by comparison with early editions or manuscripts, a practice of which Goethe expressed disapproval. Grumach's criticism was that what had not been undertaken at the time, namely a real recension of the text, had not been undertaken by the Weimar edition either (1887-1919), in that the edition simply followed the principles valid for the "final

revised edition", and departed from the text only in the case of obvious corruptions, declaring "o n e edition sacrosanct and regarding all other witnesses . . . as contributory only". "The task of the critical editor", according to Grumach's 1950 phrasing, "can only be to produce the best text", that is to say, "to examine the text conscientiously and to the best of one's ability, taking into account all available textual evidence and all factors affecting transmission and development of the text, in order to produce the version of the text which gives adequate expression to the intention of the author."[2]

The realisation of Grumach's concept can be examined in the two volumes which appeared under his supervision, containing text and apparatus of Goethe's epic poems.[3] The transmission and establishment of the text of two poems is presented here in abridged form, as an example of a German edition approaching Greg's concept. The most important authoritative documents for Goethe's epic Reineke Fuchs (4312 hexameters in 12 cantos) are:

• E First edition in Goethe's Neue Schriften, vol. 2, 1794. Printer's copy not extant. Goethe did not read proof. At least 57 misprints (in Grumach's estimation), some of which are corrected in a printed errata list probably made at Goethe's request.
• A Works, vol. 10, 1808. The printer's copy was a revised exemplar of E, probably with considerable alterations in punctuation and reshaping of a number of lines by Goethe. There is no evidence that Goethe read proof. In Grumach's estimation, 54 misprints in the text.
• B Works, vol. 11, 1817. The printer's copy was a revised exemplar of A; Goethe had altered wording and punctuation in the first four and in the last (12th) canto. Goethe is not known to have read proof. 25 new misprints in the text.
• C Works, final revised edition, vol. 40, 1830. Set from an exemplar of B, corrected by Göttling according to Goethe's directives. No alterations in the text demonstrably attributable to Goethe. Goethe did not read proof.

The text of the Academy edition (designated hereafter as the edited text) was in this case not based on the first printing in E, but, following Grumach's general concept, on the text as printed in A, being Goethe's last complete revision of the work (B having been revised by Goethe only in five out of twelve cantos). Departures from the base text A occurred (1) in order to adopt such variants in wording and punctuation from B as Grumach attributed to the author, that is to say, in order to adopt the last certain authorial alterations; (2) in order to correct the 54 misprints in A, for which purpose the reading of the first edition E was introduced, or, if E was also corrupt, the reading of the nearest witness in chronological sequence, B or C; (3) in order to standardise the spacing.

Goethe's epic Hermann und Dorothea, with approximately 2000 lines, displays a rather more complicated textual history:

• H3 Scribal copy, written 1796/97, with several layers of corrections by Goethe.
• H4 Lost printer's copy of E, probably a fair copy of H3.
• H5 Subsequent list of corrections to H4, disregarded by E in three places.
• E First edition, separate publication, 1797. Set from H4H5 and from a further lost list of corrections to H4. E is a correct text without misprints!
• N Pirate reprint of a reprint of E, with 241 corruptions of the text.
• A Works, vol. 10, 1808. Set from a copy of N as worked over by Goethe. 141 misprints were reproduced; in two places the corrupt wording was varied; in two places the corrupt punctuation was altered; in the remaining cases the original reading was restored; 13 additional new misprints. No evidence that Goethe read proof.
• B Works, vol. 11, 1817. Set from A. Some new misprints; variants in orthography and punctuation, indicating a general scanning of the text but not thorough corrections by Goethe. No evidence that Goethe read proof.
• C Works, final revised edition, vol. 40, 1830. Set from an exemplar of B, corrected by Göttling according to Goethe's directives. 5 new misprints. Goethe did not read proof.

In the establishment of the text, the first edition was regarded as superseded by the revised version A. The last complete revision had to serve as base for the edited text, that is to say, A. Since the 141 corruptions which had penetrated from the pirate edition could not be regarded as intended by Goethe, but only passively authorised, the edited text reverted in these instances to the correct text of E; the two places where the corrupt punctuation had led to an alteration (probably by the compositor), and the 13 places newly corrupted in A were handled similarly. In the two places, however, where the corruption of N had led the author to make a correction, this correction was left in the text as representing the final intention of the author. Moreover, those three variants were adopted from H5 which had not passed over into E. Thus no variants were adopted from the versions subsequent to A, because none of these exhibited individual alterations which could be proved to go back to the author.

The extent to which this editorial procedure coincides with the principles of Sir Walter Greg is evident: each individual variant is investigated to see whether it originates from the author. If this cannot be made to seem probable, the variant is not adopted in the text. In this way the edited text appears as an eclectic (contaminated) text. The fact that the text is not edited on the basis of the witness containing it in its earliest published form (in analogy to the concept of "copy-text editing") but instead on that of the last version fully revised by the author has to do with the fact that German editorial procedure does not distinguish in so marked a way between accidentals and substantives as does Anglo-American practice.[4]

This establishment of text as carried out by Grumach can be taken as typical of German editing in the 1950's. But for reasons which in this case are not known to me, internal difficulties arose. The guidelines which Grumach had followed were replaced by new "basic principles".[5] These formed the core of a new concept which has by no means gone unchallenged, but which has been winning more and more support from editors, it seems to me, since it gradually became known. Implicitly, the stipulations include among other things some indirect criticism of the methods of establishing text which had previously been customary, in particular of Grumach's procedure. Inasmuch as Greg's concept and that of Grumach coincide, there is also implicit criticism of the former, when applied to conditions of transmission such as those typically obtaining for writers of German literature since the 18th century. I shall take up those points which are essential in the present context and attempt to develop the concept somewhat further. Matters of principle are raised above all by the following

questions: (1) Is contamination admissible, that is to say, may the editor transfer variants from one authorised version to another, as happened in the Academy edition and as the Greg-Bowers principles require? (2) What is meant by the (final or non-final) intention of the author, and how may it be determined? (3) What should be regarded as textual error?

I.
The Concept of Version ("Fassung")

"Eclectic editing", the establishing of a text from more than one authoritative document, comprises two types of multiple authority which are essentially different from one another, although they may occur in combination. In the first type, the missing original is represented by several radiating texts, from which their common original is to be reconstructed through the adapted application of the stemma rules of classical editing.[6] I too approve this procedure. What is under dispute here is the second type of multiple authority: only in this case do I speak of contamination. Here one eclectic text is produced from authoritative documents which differ from one another in essence through the intervention of the author. This eclectic text is supposed, among other things, to realise the final intention of the author, as far as this is possible on the basis of extant documents. In the first type of multiple authority the textual differences are exclusively errors caused by the process of transmission; in the second, there is an added difference between versions which is intended by the author. Texts with authorial variation I designate as different versions ("Fassungen"). In the most extreme case a version is constituted by a single variant. A holograph with one alteration which does not simply correct an error thus represents two versions of the text. Though it is unusual to express oneself in this way, it is not possible to give a useful definition of the concept in any other terms. The significance of the definition will become apparent immediately.

The main difficulty in realising Greg and Grumach's undertaking, namely in establishing a text representing as reliably as possible the demonstrably final wish of the author, lies in separating from one another transmissional and authorial variants. In order to achieve this, Greg set up the well-known rules: "An editor should in every case of variation ask himself (1) whether the original reading is one that can reasonably be attributed to the author, and (2) whether the later

reading is one that the author can reasonably be supposed to have substituted for the former. If the answer to the first question is negative, then the later reading should be accepted as at least possibly an authoritative correction. . . . If the answer to (1) is affirmative and the answer to (2) is negative, the original reading should be retained. If the answers to both questions are affirmative, then the later reading should be presumed to be due to revision and admitted into the text. . . . It will be observed that one implication of this procedure is that a later variant that is either completely indifferent or manifestly inferior, or for the substitution of which no motive can be suggested, should be treated as fortuitous and refused admission to the text."[7]

Characteristic of this procedure is that the same set of questions is applied to each individual reading, and it is left to the discerning judgement of the editor to classify it. There can be no doubt but that what results is assured of the maximum authority with regard to the sum of all its elements.[8] The question is, however, whether the sum of authoritative readings yields an authoritative text. Should not the authority of a text be considered to extend equally to the texture of the text, to the relationship of its elements to one another and to the whole, and therefore to what constitutes a text as a text, to what makes it into a particular version?

As far as I know, the methods of textual constitution under discussion take this basic aspect of the version into consideration only in special cases: "In some literary works it is generally recognized that a revision may be so thoroughgoing — so motivated throughout by the author's altered political, social, or artistic concepts"[9] — that the variants cannot be transferred and the editing of more than one version in parallel texts or in separate editions is required, because the later version represents an entirely new creative act, and demanded in a sense a new manuscript. As examples the different versions of Wordsworth's Prelude are cited, or Henry James's revision of his early works

Roderick Hudson and The Portrait of a Lady. "There are two 'final' versions." "Each of these texts deserves to stand in its own right as a separate text."[10] I assume that the differentiation is made not merely for the sake of avoiding an overflow of variants in the textual apparatus, that is to say, for practical reasons, but as a response, in these cases, to an altered authorial conception of the work, and thus for reasons of theory and principle.

The question then arises as to whether it is in fact possible to mark a borderline between the two differently treated cases, whether it is practicable to make the distinction in every instance. One could not, for instance, decisively distinguish two versions by quantitative criteria, which might demand that the variants should exceed a certain number, or that a certain time must have elapsed before the revision. In the line "Dass endlich zerbräche das kühle Haupt", in the last version of Trakl's poem Passion (Agony), the manuscript has the adjective "kühle" (cool) for deleted "glühende" (glowing), but one cannot maintain that this alteration is more far-reaching than when in the same manuscript the "Purpurne Wolke" (purple cloud) is replaced by "Goldene Wolke" (golden cloud).[11]

The first stanza of Hölderlin's Feiertag hymn describes parabolically the condition after a storm. In the prose draft of the poem we read "vom erquikenden Reegen des Himmels"; in the metrical version "von des Himmels erfreuendem Reegen". At first sight one might not raise objections if an editor or commentator considered the two readings ("erquikenden / erfreuendem Reegen") simply in terms of stylistic improvement as the replacement of a commonplace expression by a clearly metaphorical adjective. Yet related within the text the matter looks different:

• 1 Wie wenn am Feiertage, das Feld zu sehn
• 2 Ein Landmann geht, des a) Abends, b) Morgens, wenn
• 3 Aus heisser a) Luft b) Nacht die kühlenden Blize fielen
• 4 a) Den ganzen Tag b) Die ganze Zeit und fern noch a) hallet b) tönet der Donner,
• 7 Und von des Himmels a) erquikendem b) erfreuendem Reegen
• 8 Der Weinstock trauft

One might still regard the variants in l. 4 "hallet / tönet" in this manner, namely as synonyms, and similarly, without regard to the context, "Den ganzen Tag / Die ganze Zeit" in l. 4. But related within the text it becomes apparent that this is just as impossible as with l. 3 "Luft / Nacht", and that the same difference exists as between the variants in l. 2 "des Abends / des Morgens". Here it is impossible to regard the variants as synonymously varying expressions. The variation emerges as a contrast in meaning, and therefore in intention, and thus it is determined that in l. 4 "Tag / Zeit" are not synonymous; rather, "Tag" 4a contrasts with "Nacht" 3b. These variants do not only occur between the prose draft and the metrical version, but also internally in the latter.[12] The draft and the metrical version are directly consecutive in the manuscript, and were presumably written down in direct chronological sequence. The variation attests a differing concept of time within the one manuscript. Therefore, contamination between the variants of the two versions is out of the question, Instead of saying merely that the variants correspond to a new concept, however, I can also say that there is a certain relatedness between them, that they result from a relatedness, the relatedness of a version.

In one quite specific case contamination in establishing a text is probably the only correct procedure. This is when it can be proved that the author did not make alterations which bore a relationship to one another and to the whole, but simply altered things here and there in isolation. But can that be proved?

In adopting the procedure of contamination the editor starts in general from the premise that the intention of the author and the concept of the work remain the same even when a revision is made, unless this assumption can be disproved by such obvious contradictions as that "Abends" and "Morgens", "Tag" and "Nacht" be synonymous. Such cases are regarded as special cases. But in the remaining cases the editor is to my mind doing in principle just what he would be doing if he were to adopt the procedure of contamination with Hölderlin, only it is less obvious. Fredson Bowers once said of a distinction of this sort, an editor who strains at a gnat may swallow camels too.

What is disturbing to me in this construction of categories is that it is not founded on any explicit theory, that it tends rather to mark as special cases those intractable instances which would in fact refute

the implicit hypothesis. The hypothesis consists in the tacit assumption that the alterations of the author do not in general result from relatedness,[13] and the assumption in turn is based on the usual absence of striking evidence to the contrary. The very opposite answers to my own general view of the essence of a literary work, but this again can be proved only in special cases, for instance when there are close syntactic links between the newly introduced textual elements, or links in rhyme or metre, or—as in the instance cited (Hölderlin)—clear semantic relationships. But what becomes obvious only in such instances is to my mind almost always the case. My conception rests on the linguistic idea of the text as a complex of elements which form a system of signs, both denoting (signifiant) and denoted (signifié).[14] That it is a system means that the work consists not of its elements but of the relationships between them. However, the clear relationships between the elements form only a part of the relevant relationships. Relationships exist likewise on the phonological level of the text, on the lexical level, etc.

Seen in this way, a version is a specific system of linguistic sings,

functioning within and without, and authorial revisions transform it into another system. As I said, in principle a new version comes into existence through a single variant. Since a text, as text, does not in fact consist of elements but of the relationships between them, variation at one point has an effect on invariant sections of the text. In considering different versions one must therefore not confine one's attention to the variants. This is most clearly exemplified when the title of a work is altered. Fundamentally, therefore, whether the variants are numerous and of far-reaching effect is not a necessary condition for the constitution of a version.[15] A new version implies a new intention.

II.
The Problem of Authorial Intention

The application of the Greg-Bowers copy-text theory requires the interchangeability of variants from different versions. This requirement, as far as I can see, rests on the assumption that the alterations made by the author are isolated improvements within a concept which remains constant, unless the opposite can be proved, which it rarely can. This assumption in turn is usually accompanied by a particular notion of the mutations of a literary work; in German studies this notion is rooted in Goethe's understanding of himself as man and poet. As was described in the introductory section, in establishing Goethe's texts from the final revised edition, Grumach had in each instance adopted only substantive alterations attributable directly to

Goethe, not, however, the alterations undertaken by Goethe's assistants under his surveillance—not far-reaching standardisation of punctuation, orthography and word forms. He regarded these as alien to Goethe's text, often detrimental to Goethe's real intentions; that they were retained in the Weimar edition (1887 ff.) is said to have been the result of "false piety". At this point, criticism of Grumach's edition set in. Friedrich Beissner founded his objections on "the right of the author to his own work", and declared that for a complete edition the principle of the Weimar edition had been unquestionably right in giving favourable consideration to the final revised edition including the standardisation, because Goethe wanted it, and sanctioned the procedure of his assistants, and the result. An editor, however, would be obliged to carry out Goethe's final intentions better than his assistants did at that time; he would have to standardise according to more refined methods and at the same time "preserve what was distinctively Goethe's". He would only be entitled to reverse alterations which "could be proved to be in opposition to Goethe's real intentions and which had been overlooked by him."[16]

The establishment of a text which appeals ultimately to the intention of the author may easily give rise to controversy. To the process of standardisation mentioned, which reflected in the main the scribal habits of the 1820's, Goethe also subjected works which had come into existence 50 years earlier with quite different graphical characteristics, works which had been the foundation of his literary fame. Thus according to his final intention Goethe wished to appear before his readers not out-moded and old-fashioned, but in modern, contemporary orthography. Would it not be right for an editor citing the intention of the author in support of his edition, believing himself obliged to carry out what the author himself no longer carried out, to derive from this even the obligation to transpose Goethe's works into modern spelling, "while with utmost sensitivity respecting and preserving what is distinctively Goethe's"? Were this to happen, a basic principle of the historical-critical edition would be injured by the principle of the final intention.

The axiomatic preference for the final intention of the author as against his earlier intentions can be seen in two ways. Either it is based on the teleological notion that the work itself has a goal, or that the author envisages one, and that in the course of time he brings his work nearer and nearer to it; correlated with the work is a concept

which the writer's "improvements" can realise with increasing refinement and precision. (Therefore it is emphasised that only a change in concept hinders the application of the copy-text theory.) Or it is based on the related morphological notion that the work is an organism, that its mutations are a "development", an "organic growth", a continual process of ripening, from the seed perhaps even to perfect maturity. This was how Goethe saw his life's work. And this was how the editors of the Weimar edition saw Goethe's final revised edition: "By nature [!] the last edition represents the summit, the conclusion."[17]

Moreover another concept is adduced, which in my opinion signifies an old confusion of the textual and the legal spheres, namely when that same edition is described as "Goethe's testamentary disposition", as "his bequest", and when the editors feel themselves to be his literary executors, and believe that they are commissioned by him to carry out what he wished to undertake but did not himself complete. If one wanted to pursue this editorial principle of final authorial intention to its logical conclusion, it would lead in some cases to the annihilation of the object of the edition through the edition: even today one would have to destroy The Aeneid. For Virgil demanded of his friends Marius and Tucca that they should burn the work which at his death was extrinsically complete, intrinsically unfinished; it was to be destroyed for purely artistic reasons, because it did not fulfill his standards. Augustus ordered them to publish the poem.[18] Franz Kafka demanded of his executor Max Brod that he should burn his unprinted writings unread.[19] Instead he published them. This time there was no emperor. But only the executor, not the editor, can be set at odds by such an expression of the author's wishes; the editor's philological task, here as elsewhere, can only be to interpret extant documents and accompanying circumstances as historic facts. In my opinion he has to deal with the intentions of the author not as an executor, but only as a historian, and he should regard them not as binding directives for editorial decisions, but as historical phenomena.

And I believe that this is the way in which texts should be regarded altogether, in their relationships to one another, to their originators, and to the conditions of their genesis and influence. I see the following parts, or rather aspects, of the history of a work: (1) the history of the text, that is to say, the history of its emergence and the alterations

made to it by the author; (2) the history of its influence and reception, which begins as soon as the work, or parts of it, become known to other people; (3) the history of transmission, that is to say, the process following on from the textual history, where, even under the eyes of the author, the work begins to be affected by those intentional and unintentional textual alterations which accompany its dispersion. The aspects overlap at times; unauthorised "facsimile" reprints ("Doppeldrucke") and pirate editions with their distortions may provide the basic copy for an authorised revision of the text. Of importance for our discussion is above all the interdependence of factors of textual history and of influence and reception. Apart from rare exceptions, it is safe to assume that authors want to be read, and that in writing they think of a public. For this reason an author not only has an effect on his readers, he is in turn affected by their reaction. It is not possible (or only rarely, in exceptional cases), when a work is revised, to give a detailed account of the extent to which the reception of the first version, a change in society, a change in the author himself and in his relationship to his environment, a different incentive or purpose in publication, may be involved in the revision, and this holds all the more true since right from the beginning, before he even thought of writing, the author was exposed to this play of forces from all sides. What is termed the intention of the author is an undetachable part of these forces, and therefore seems equally ill-suited as a criterion for editorial decisions or as a criterion for the evaluation of literary works. Only the textual history is within the editor's reach: notes, extracts, drafts, when they have survived, and then the versions in chronological sequence, a diachronic succession of discrete semiotic systems. The totality of the versions yields the diachronic pattern of the work. Each individual version yields a synchronic linguistic pattern.

If one imagines the textual history in the shape of a 3-dimensional cylinder standing upright, then the different versions are horizontal planes perpendicular to the axis of the cylinder. The purpose of the historical-critical edition (apart from the necessary correction of mutilated text) is to create an appropriate reproduction of this cylinder, that is to say, of the complete textual history; while the purpose of a critical edition is to reproduce a particular plane, that is to say, an individual version. Contamination would mean the projection of one plane onto another. In the historical-critical edition the editor selects one version (or, when there are substantial differences between the versions, more than one) for reproductions as the text in his edition, and he

presents the remaining textual history (or possibly all of it) in the apparatus.[20] From a historical point of view the different versions are in theory of equal value. Each represents a semiotic system which was valid at a specific time, which the author later rejected, because he for some reason no longer found it adequate, in favour of another version which matched his new intention. To explain the alterations by saying that one assumes the author's intentions are still the same but that he was not previously able to realise them so well, is to declare him a bungler. Whether the author regards the alterations as an improvement of his work, or as an enhancement of its literary qualities, whether or not he declares the earlier version invalidated, or condemns it—for the historian, for the editor, the alterations mean an adaptation of the work to suit the altered circumstances, ideas and purposes of the author. For the editor there is no "best version". In selecting the text to be edited he is not bound by the final intention of the author. He may, for instance, defend the young Goethe's works against the revision of the 75-year-old, of whom it can be proved that he in part no longer understood, or no longer wanted to understand, the creations of his youth. The editor of Ernst Jünger's works will probably reject the alterations which the author made to his earlier works after the collapse of the Third Reich, and edit the original versions. In specific cases the choice of version for the edited text may depend on very different factors; the history of influence and reception may be decisive, or again the degree of corruption or purity of a particular version. As long as the editor sees his function as that of a historian, he has a wide range of freedom in the selection of the version for the edited text, but this version he must reproduce without contamination.

The establishment of a text oriented to the intentions of the author is confronted with considerable difficulties when it becomes apparent from the variants that the magnetic needle of the author's wishes is quivering in the field of non-aesthetic forces. This leads to characteristic problems. The printer's copies for the first edition of some of Nathaniel Hawthorne's novels have survived, fair copy autographs. The manuscripts of The Blithedale Romance (1852) and The Marble Faun (1860) display traces of the influence of Hawthorne's wife Sophia. Crosses in the manuscript of The Marble Faun, as the

editor tells us, mark places which Sophia noted in order to suggest to her husband an alteration (in style or content). Usually the author complied, as the corrections show. In three marked places, however, he resisted. One of these places, with the phrase "vinous enjoyment", is characteristic of Sophia's temperance convictions.[21] In the manuscript of The Blithedale Romance Hawthorne made excisions only in three places; two are concerned with liquor and one with sexuality. It is not possible to ascertain, says the editor, to whose initiative the deletions are to be traced, but it seems very probable that they "were made in deference to the sensibilities of Hawthorne's wife. . . . Although the deletions undoubtedly represent his final intentions, the causes behind these intentions are suspect." "The reasons could not have been literary."[22] Therefore the deleted passages have been re-adopted in the critical text. It seems to me that in the editor's decisions the difference which obtains in the case of The Marble Faun, where Hawthorne resisted the influence of his wife, does not find adequate expression, and for methodological reasons I find the idea perplexing, that the editor should feel obliged not only to make inferences about the final intentions of an author, but also about the causes behind these intentions. I cannot regard the psyche of the author and its analysis as a substantial foundation for editing.

All cases of actual and suspected self-imposed censorship give rise to this kind of analysis. Fredson Bowers, as editor of Stephen Crane's Maggie (1893), did not by any means put a veil over the difficulties involved in such distinctions; he placed them clearly in the foreground on several occasions.[23] The first version of the novel could only be printed privately under a pseudonym. When Crane found a publisher for it three years later, the publisher demanded an attenuation of the drastic language, that is to say, expurgation "to remove most of the profanity and perhaps some of the sordid detail characteristic of the original version"[24] (in Crane's own phrase, "words which hurt"). Problems arise, because it is clear that Crane voluntarily undertook further corrections and in some instances it is not clear to which category the alterations belong. This is the case with three deletions

in chapter 17 and with some modifications, which were carried out either "under the impression that they would be required by the general agreement . . . about the nature of the revisions",[25] or for purely artistic reasons, for the "softening of satire or of shocking detail".[26] The editor feels obliged to come to a decision on this score in each case, and to include in the edited text only those departures from the first version which match "the author's final and uninfluenced artistic intentions".[27] The investigation of the nature of the deletions in the 17th chapter constitutes a highly impressive interpretation of the two versions of this chapter ("Textual Introduction," pp. lxxviixci), and leads to the conclusion that there are no grounds for assuming anything other than a deletion made by Crane deliberately for literary reasons, in order to resolve an ambiguity. One might of course ask whether an investigation of this sort should be the concern of an editor or of the textual critic, or rather of the literary critic.[28] Greg's principles, in particular the aim of clarifying "whether the later reading is one that the author can reasonably be supposed to have substituted for the former,"[29] and most particularly this "reasonably", require the editor to undertake ad hoc investigations of this nature whenever an alteration is suspect. However, would it not ultimately put him under the obligation not only of examining the variants, say, stylistically, but also of interpreting, with regard to the alterations, the original and the revised version of the text, taking into consideration all its relevant dimensions, and would this kind of interpretation not lead directly into the hermeneutic circle, inasmuch as the revised version represents both the basis and the goal of the investigation? It is indeed notable that scattered through the comments of the editor on the motives behind certain of Crane's softenings are phrases such as "could be", "it is not certain", "it is possible", "probably".[30] The reason for this does not by any means lie in a deficiency in the editor's powers of analysis, but in the thing itself: if one has to reckon with purely literary and purely censorial motives for alteration, one must also reckon with a combination of the two. Firstly, in anticipation of the character of the expected censorship, Crane could be led to undertake alterations which also had literary value in the context of the new version. Secondly, because of the systematic character of the work, purely censorial alterations sparked off further alterations, determined at this stage by literary considerations. Again in consequence of the systemic character of the work, the contamination of the two historical versions in the edited text gives rise to a third version. Though the editor may indeed give a rational account of his decision at each point on the basis of the documents, nevertheless to aim to produce the ideal text which Crane would have produced in 1896 if the publisher had left him complete freedom[31] is to my mind just as unhistorical as the question of how the first World War or the history of the United States would have developed if Germany had not caused the USA to enter the war in 1917 by unlimited submarine combat. The nonspecific form of censorship described above is one of the historical conditions under which Crane wrote the second version of Maggie and made it function. From the text which arose in this way it is not possible to subtract these forces and influences, in order to obtain a text of the author's own. Indeed I regard the "uninfluenced artistic intentions" of the author as something which exists only in terms of aesthetic abstraction. Between influences on the author and influences on the text are all manner of transitions.

It is in the nature of certain areas of literature that, apart from aesthetic purposes, obligations exist towards practical purposes and diverse forms of non-literary ideology; this is the case with the so-called utilitarian forms such as the essay, the pamphlet, memoirs, originally with all rhetorical forms and with André Jolles' "Einfache Formen", fable, didactic poem and satire, the mediaeval mystery play and legend, Jesuit drama and Brecht's plays of instruction. Klopstock still does not recognise any autonomous poetry of quality. His poetic message is bound up with his religious convictions and obligations. The highest form of poetry, the Christian epic, and religious poetry altogether, is proclamation of faith. "The plan of the revelation is its prime rule", he says of the "sacred poem". Nevertheless, in his major work, Der Messias, he undertakes alterations and deletions for other reasons too, for instance in order to prevent the work from becoming the object of theological disputes detrimental to it.[32] For me it is inconceivable that editors should be allowed to reverse such alterations on the grounds that they were made for reasons neither religious nor poetic. In

German literature it was only during Klopstock's lifetime and later that poetry laid claim to an autonomous area of its own, and even then it naturally retained links with the rest of reality.

The editor oriented to the artistic intentions of the author is confronted also with those puzzling situations which James Thorpe treats in the second part of his essay, "The Aesthetics of Textual Criticism",[33] asking "what constitutes the integrity of the work of art?" Often the editors of magazines intervene not only in the accidentals of contributions. But even authors who have protested against such changes often adopt them when they publish their periodical contributions in book form, and seem to give them some kind of authority. How is the editor to interpret this behaviour on the part of the author? "Often we cannot be sure whether he makes suggested changes because of a compliant disposition, whether he allows editorial alterations to stand in later editions out of laziness, whether he reverts to earlier readings out of pertinacity, or whether there is reasoned conviction in support of his action".[34] Thorpe comes to the conclusion that the intentions which amount to "the integrity of the work of art" include "those intentions which are the author's, together with those others of which he approves or in which he acquiesces".[35]

I think that it is the difficulty, indeed the impossibility, of obtaining a text attributable exclusively to the author, when conditions are really complicated, which led Thorpe towards the recognition of an aggregate of alien influences. In this he adopts a position very close to that of current German studies being described here.

III.
Textual Error and Textual Fault ("Textfehler")

Our great sensitivity to contamination will be countered from the viewpoint of the Greg principles in the following way:—in transposing the last reliably authoritative substantive variations into the copy-text, which is binding only for the accidentals, one is basically not mixing different versions, but simply cleansing the last version of non-authorial intervention. I am prepared to agree that under quite specific circumstances this is indeed the best way to produce the correct authorial text; namely, if one has as copy-text an autograph printer's copy, if the author can be shown to have read the proofs only cursorily, and directly after the production of the copy, and finally if he can be

shown to have exerted no influence on subsequent editions. These conditions are fulfilled, for instance, in the case of some of the novels of Nathaniel Hawthorne (such as The Blithedale Romance). In a case like this it is my opinion too that one produces the text closest to the author by examining the authority of every single deviation of the printed text from the printer's copy (with the reservations mentioned above regarding the excisions). Only when these conditions are fulfilled does the procedure not conflict with my concept of the versions of a literary work: in such a case one can indeed take it as likely that the author made his corrections and revisions to the proofs on the basis of the idea of the text which was still fresh in his mind from the manuscript, and not on the basis of the text set and altered by the compositor.

Where these conditions are not fulfilled, the Greg establishment of text runs in danger of contamination. What I mean can best be demonstrated by an example. Because it is a matter of relationships within a work, it is necessary to present the relevant passages within the context of the work. For this demonstration a ballad by C. F. Meyer is particularly suitable, for which ten textual witnesses have survived from the years 1871-1892. I discount the earliest manuscript because the second document (designated here as H1) has been revised to such an extent that it and the subsequent versions cannot easily be related to the first. The nine documents H1-D9, on the other hand, can be registered in synoptic sequence in such a way that the alterations of each as against the preceding one can be read off; only text which varies from that of the directly preceding version is registered in each case. Text deleted in the manuscripts is in [square brackets]. Certain variants in punctuation are given in the margin; orthographic variants are listed at the end, so as not to encumber the presentation with them. The nine authoritative documents are as follows:

       

H1	Manuscript written by C. F. Meyer's sister, dated "January 1873".
H2	Manuscript written by Meyer's sister. Revision of H1. Printer's copy for D3.
D3	First printing in an almanac, 1873. Deviations from H2 only in punctuation (subtitle and l. 19) and orthography. No misprints.
D4	Authorised reprint in an anthology, 1876. Deviations from D3 in wording (subtitle and l. 33), and in punctuation, and graphical changes. Meyer to the editors: "I wish to read proof myself". 8 probable misprints.

 

D5-D9

C. F. Meyer, Gedichte, 1st-5th editions, 1882-1892. D5 is a revision of D4. D6-D9 alter only punctuation (D8 l. 20; D9 l. 13) and orthograph. In each case Meyer worked over the five editions of the collection thoroughly, and read the proofs, particularly of the first three, carefully.

 

D5-D9 have no stanza division. At line 11, H1 displays alternate readings ("Alternativ-Varianten"), neither of which is deleted; both therefore have the potential for subsequent adoption. In punctuation and orthography, the synoptical presentation follows the first printing D3. Orthographic variants not registered in the synopsis are: title, 22, 33 Komturs, Komturei D9 6 Küssnach D7-D9 6 dreissig D4D5D8D9 31 rot D5-D9 35 Heimschritt D7-D9

The ballad is based on a saga from the Swiss Wars of the Reformation. From Küsnacht, a village on the Lake of Zurich, the Grand Commander Schmid, a Knight of the Order of St. John, rode out with other men from Küsnacht to help his friend, the Zurich reformer Zwingli, in battle. At this point the ballad begins. It describes the anxious waiting of those who remained behind, and the return alone of the fallen Commander's charger. The charger is said to have swum across the lake by night and to have entered the stables among the buildings on the Order's estate. At this point (l.33) the text of the first three witnesses has: In Meyer's collected poems on the other hand all editions have: For someone who knows the scene of the ballad as the poet did (Meyer lived for several years in Küsnacht), the original version has particular significance; a late Gothic choir of unusual height still rises high above the nave of the church, as it did in the Grand Commander's time. To leave no doubt about this detail, Meyer emphasised the C of "Chor" by underlining it in the printer's copy H2, leaving no room for doubt in the mind of the compositor. But later the mistake seems to have been made nevertheless, when the compositor of the anthology D4 set the text (presumably from D3). This compositor is probably to be blamed for seven further misprints in the ballad, and also for errors in the 14 other poems and in the prose of C. F. Meyer printed in this anthology, some of them serious distortions of the text. Meyer's wish to read proof himself was probably not fulfilled. The phrase "Thurm und Thor" (l. 33) lacks vision, but it can easily be explained as the result of confusion of the letters C and T by the compositor, as the substitution of the smooth but trivial expression "Thurm und Thor" for the compelling phrase "Thurm und Chor". I am convinced that we have to do with a compositor's error in D4 at this point and in the loss of the apostrophe indicating the preterite in ll. 1, 25, 32, 34. When Meyer revised the ballad six years later for his collected poems (D5), he clearly used the version D4 and let the phrase "Thurm und Thor" stand; similarly in all following editions. When I was editing the text of the collected poems twelve years ago, I adopted the view that this could only be a matter of a misprint taken over from D4, which as an alliterative formula benumbed the thinking and the perspicuity of the author and his helpers, and was thus able to elude the very thorough corrections of the first three issues of the collection (D5-D7). I therefore put the original phrase "Thurm und Chor" from H1H2D3 into the edited text.

I still consider "Thor" in D4 to be a misprint, but in spite of this I consider my emendation to D5-D9 to have been wrong. I now believe that "Thor" corresponded to Meyer's intentions when he produced the version D5 from D4, and I believe that in contrast to the versions H1H2D3 his intention shifted from the optical, say, to the acoustic, from the image "Thurm und Chor" to the alliteration "Thurm und Thor." This alliteration is very well-suited to the popular tone and archaic language (l.27) of a ballad treating of legendary material. Because of the somewhat singular architectonic conditions, the optical image is an artistic touch which can scarcely be appreciated. There is no indication that Meyer rejected the variant "Thor". But one cannot say whether he recognised the misprint as such and consciously adopted the variant in the text of D5. Be that as it may, I believe that D4's misprint influenced the alteration of intention observable throughout D5, and that it is given authority by this new intention; thus in D5-D9 it can no longer be regarded as a textual error (as it was in D4).

The credibility of this assumption is strengthened by a comparison with the other instances of damage to the text in D4, to which the author responded by adopting some in D5 and by varying others. In ll.14a, 14b the postpositional subject has been mistakenly severed by a full stop in D4. D5 does not alter the faulty punctuation,[36] but instead the context; presumably the concluding stop in l. 14 drew Meyer's attention to the expendability of ll. 14a, 14b. We do not have to assume that the misprint in D4 caused an oversight in D5 (the omission of two lines), but that Meyer's reaction to the disfigurement in D4 was to retain what had originated as a misprint while altering the context in D5—parallel to l. 33 "Thurm und Thor". In ll. 32-34 we find further misprints in D4, whereby the verb forms "Verkündet'" and "Ragt'" lose the apostrophe, the only preterital sign. Whereas in

l. 1 "legt'" and l. 25 "starrt'" Meyer reversed the same error, in ll. 32 and 34 he lets the present forms stand, so that presumably the misprint again causes a change in the system, this time in the temporal structure of the poem. In H2D3 the whole poem is in the preterite, with the exception of the direct speech. D4 certainly has four erroneous forms of the present. D5 on the other hand has preterite everywhere except in ll. 30-34, and this is deliberate as the change of tense in l. 30 "überquillt" D5 for "überquoll" D4 shows; after the speech of the boy in l. 28 the narrative present is introduced. Thus it is apparent that in producing the version D5 from the text of D4, Meyer reversed only two (ll. 1, 25) of the eight disfigurements in D4. In the remaining six cases he either adopted the erroneous reading unaltered (l.33), or he changed it again (ll. 20, 21), or he adopted the erroneous reading and altered the context (ll. 14, 32, 34).

If I simplify the textual conditions in Meyer's ballad to schematic form, I can say that the relationship between the almanac printing D3 and the anthology copy D4 is analogous to that between an autograph manuscript and a faulty printing set from it, of which the author either did not read the proofs at all or only in a very cursory manner.[37] Such special conditions obtain to a considerable degree for Hawthorne's The Blithedale Romance, as was mentioned above. Following the Anglo-American usage, I can designate D3D4 as one version, and can call this the first version, since I discount H1 2. In this way D3 could be said to consist of the relationships between the elements A-H. In D4 the misprints F' G' H' replace F G H. If the editor wants to adopt this first version (D3D4) in the edited text, he will be well-advised to reverse D4's misprints F' G' H' in order to restore the potential of the relationships between the elements E-F, F-G, etc., established by the author.

     

1st version	{D3A — . . . — E — F — G — H {D4 A — . . . — E — F' — G' — H'
2nd version	D5 A — . . . — e — F' — G — h
Contamination	D3/D5 A — . . . — e — F — G — h

But for the production of the second version (D5) the special conditions (of The Blithedale Romance) no longer obtain; it is a revision on the basis of the faulty text found in D4. As we said, the author produced D5 partly by using original elements from D3 which are also in D4 (A . . . D), partly by altering the original text (e), partly by adopting damaged text (F'), partly by reversing mistakes (G), partly by altering such readings (h). Instead of the relationships of the first version which were intended in D3 and disturbed in D4, we have other relationships.[38] If the editor wishes to adopt the second version (D5) in the edited text, he must give the text of D5 which the author produced (possibly with emendation of such new misprints as may demonstrably be attributed to D5—of which more will be said shortly). On no account may he reverse in D5 the misprints observed in D4, as the Greg-Bowers principles demand; if he did this he would make an arbitrary sychronisation of non-synchronic elements; he would produce from D3 and D5 a contaminated version unknown to the author (the bottom line in the scheme), a version which would not have the potential of the relationships as established by the author in D5, but some other instead. The reconstruction would be in danger of becoming a construction alien to the author.

The example of Meyer's ballad, and also the example from Klopstock's epic quoted in footnote 15, and considerations of a general nature, lead us to the paramount understanding that for the production of a new version the source of the elements which the author uses in the text is immaterial; it does not matter whether the variants are original or extraneous, misprints (as we shall see, there are misprints and misprints) or variants introduced by a publisher's editor. In other words, for the elements of a text to be authoritative, and thus for the text itself to be authorised, it is not necessary to assume that the author recognised extraneous elements included in the text (compositor's errors and editorial intervention) as such; he may have regarded them as original text. The necessary condition for our establishment of text is only that he should have registered the readings in question. That this is the case is sometimes obvious (for syntactic-semantic reasons, etc.), and sometimes it can be demonstrated by interpretation (as in some cases in Meyer's ballad); for the remainder (for instance with other texts by the same author and by other authors), it is an assumption which represents a generalisation made on the basis of the demonstrable cases.

This is the root of the matter—here we have the cause of the difference between the establishment of text based on versions on the

one hand, and on contamination on the other. Both work from a certain assumption in the undemonstrable cases: corresponding to our uniform view of literary works as semiotic systems we assume that the demonstrable and the undemonstrable cases are fundamentally the same; that is to say, we assume that the author took note of individual readings in the text which formed the basis of the revision (unless there are indications of special conditions as with Hawthorne's Blithedale Romance). The contaminating edition, however, if I have understood it rightly, assumes rather that the author not only ignored the misprint as such, but also the whole of the passage in question including wording, etc., unless the opposite can be proved; thus it is assumed that the two cases belong to two fundamentally different categories, but that the demarcation line between them coincides with the borderline between the demonstrable and the undemonstrable. I cannot imagine how one could account for this agreeable coincidence. From the methodological point of view, the circumstances are just like those prevailing in the demonstrability of discrete versions as described at the end of section I.

Since it is evident that misprints, transmissional errors, or even conscious non-authorial changes need not necessarily remain unauthorised from the "second generation" on, and do not by their original nature alone entitle the editor to emend the text,[39] it is necessary to find a specific term in order to designate those places where the editor does have the obligation to intervene in the text as transmitted. I term them textual faults ("Textfehler"). The question arises, under what circumstances is one to assume not only damage to the text, but also textual fault; what conditions must be fulfilled? In the matter of textual fault, as with contamination (the very risk of which we want to avoid), our considerable caution is not imposed on us by diffuse timidity, or by apprehensiveness about the vindication of editorial decisions, but by specific notions of the most simple literary actualities, and by our experience in editing.

First, the theory. By now it has been made clear why authorisation in our sense is not to be related to individual variants, but to versions. Greg's criteria for establishing authority, cited in section I, assure the text only of the authority of individual elements, not of the version or work. If a variant is not to be rejected in favour of the original reading, Greg demands that it should have originated with more than 50% probability from the author. Even if the substance of the version did not prohibit this procedure, there would be other

hindering factors, and it is these which make the question of textual fault so problematical. They all follow from simple insights into the modes of existence of literary art to which the theories of the Prague structuralists and their successors have called attention.

Fundamentally, there are two attitudes towards the text: the attitude of the reader, the exegete or the editor, and the attitude of the author. (There can of course be permutations.) The former is typically the attitude one adopts towards the transmitted text, probably corrupt at times, which is to be edited and interpreted. The exegete presumes that the author's text arose not by chance but of necessity, that it is unique, unrepeatable, artistically complete, a text of maximum significance. Similarly the editor searches in the transmitted text for the one authentic text, in comparison with which all else will be textual corruption. His attitude can only be that of the most faithful scholarly accuracy. For the author on the other hand a text is something to be created by selection from the semantic inventory of the language, from the quantum of synonyms. That synonyms exist for him, synonyms in the broadest sense of the word, is shown by draft manuscripts with their substitutions which often form whole skeins of variants. In the end the author has made a decision one way or the other, but he could have decided differently. For him creative writing does not mean necessity, but the possibility of variation.

Valéry remarked of the creative process: "There is no sense in consummating the word".[40] "In general I cannot take up anything I have written without thinking that I would have made something quite different out of it, had not an interference from outside" interrupted the creative process. Valéry saw no reason why he should be censured "because I presented to the public several, and sometimes even contradictory, texts of one and the same poem." On the contrary he felt tempted "to challenge poets to produce a multiplicity of variants or possible treatments for a single theme."[41] Such a situation naturally renders absurd the scholarly endeavours to find the one authentic text by considerations of probability. The claim to follow

the intention of the author in so doing means that the attitude of the author is reduced to that of the reader.

To edit the text according to the intention of the author, when the singularities of his intention are known to us only through this text, can be achieved only if the text is in a certain sense redundant, that is to say, predictable. But this condition is fulfilled, as experiments have shown, only in the case of utilitarian texts (e.g., newspaper articles), and not in the case of poetic texts, which are distinguished from other texts perhaps only by their non-redundancy. Errors can as a rule be recognised as violations of a principle. But it is an idiosyncrasy of artistic structures that they themselves transgress the rules or codes which they have set up in the text, in favour of new codes (internal disautomatisation).

Our personal experiences in editing confirm what this theory would lead us to expect. In the introductory section I described the two-fold establishment of text in Goethe's epic poems, first according to Grumach's guidelines, then according to the stricter "basic principles" of his successors (see footnote 5). The statements about the number of misprints in the 1808 edition of the Works (A) diverge quite considerably: in Grumach's view there are 19 misprints in Achilleis and 54 in Reineke Fuchs, but according to the "basic principles" there are 6 and 8 respectively; in Hermann und Dorothea A has 13 misprints according to the one, 5 according to the other. When, as an editor, one has the opportunity of pondering over a text for a decade or two, one becomes increasingly sceptical of one's own judgement of textual fault, because in the course of time one observes a tendency to acknowledge what had previously been regarded as mistakes as being within the author's range of linguistic possibilities. Therefore, on the basis of theoretical considerations and of editorial experience, we have reached a point of view concerning textual fault and authorisation which is to a certain extent in opposition to the Greg-Bowers principles.[42] Perhaps I may be allowed to take up a comparison which Fredson Bowers once made in a discussion of editorial principles before Greg. He compares his own procedure to the criminal procedure of certain Continental codes, according to which the accused is required to demonstrate his own innocence. Our procedure is more like Anglo-American jurisdiction,

with its initial presumption of innocence and the onus of proving guilt.

In conclusion I describe in outline our method of procedure, as far as it can be formulated in general terms at the present time. To begin with it is clear that in reckoning with the structure of specific versions we shall not start from the individual reading in order to define authorisation ("Autorisation"), but from the version. When certain conditions are fulfilled we attribute authorisation to a version as such, and treat individual passages as textual faults when there is sufficient evidence. Thus we mean by authorisation a provisional general quality, which may be intermittently suspended. It is assumed valid until the contrary is proved and is not conferred with the finality of the predication of authority according to Greg-Bowers. We regard as authorised those manuscripts which the author worked on himself, or which were demonstrably commissioned by him. Published texts are considered to be authorised if the author desired or approved their production, and if he influenced the text by supplying the printer's copy or by personal revision, or by revision undertaken at his request during the printing process.[43] (Where there is no authorised document, the text is established on the basis of that unauthorised document which is nearest to the lost authority.)

Textual fault ("Textfehler"), that is to say, an intermittent suspension of authorisation (in the case of unauthorised documents, an intermittent breakdown in transmission), occurs when two conditions are fulfilled: (1) when the reading in question admits of no sense in the wider contextual setting, or (particularly with modern literature) when it contradicts the logic specific to the text, the internal text structure; (2) when the results of analytical bibliography (in the broad sense of the term) confirm the suspension of authorisation. The bibliographical demonstration requires a detailed investigation of the circumstances of authorisation, such as is usual in Anglo-American studies. In the simpler cases the demonstration can be verified by lists of misprints, correction sheets, or by correspondence concerning individual passages, showing for instance that the author delegated the proof-reading of a particular sheet. There may be a whole cluster of arguments to be cited, whereby it is their convergence which is decisive.

The substantiation of the two criteria would require a more

detailed discussion than is possible here.[44] It is necessary that the criterion of text-specific logic should be supplemented by the bibliographical criterion, and vice versa. If one wanted to confine one's attention to the bibliographical criterion, the variant reading discussed above, "Thurm und Thor" in Meyer's ballad, would not only be a misprint but also a textual fault: it derives from the printed copy D4, is shown by comparison with the printer's copy D3 to be a misprint, and stands in surroundings which exhibit numerous other misprints. To the surroundings in a broader sense belong other poems by Meyer in the same anthology, which also exhibit misprints, and it must be assumed that the author could not read proof as he desired. A passage from one of the other poems provides an example of the fulfillment of both criteria. The poem entitled "Das Glöcklein" describes a dying woman's dream of death. She listens as the flocks return home in the evening from the alp, everything becomes quiet, and at the end the sound of one bell left behind rings out: The text of the first line is meaningless in the immediate context (as it happens, other documents show that "auch" is a misprint for "aus"). The text of the third line is meaningless in the broader context of the whole poem, because the homecoming is not to the "Erde", but to the eternal home, to which the "Herde" (the word occurs twice) is returning in this poem. Thus here both the bibliographical criterion and the criterion of text-specific logic are fulfilled.

The assessment of what is possible in the language of a text is subject to considerable uncertainty for the reasons of theory discussed above if for no other: the non-redundance of the work of art and its tendency to violate on certain levels the principles which it has itself established—causes which one may recognise in Goethe's observation to Eckermann, that a good poem must as a whole be rational, but in detail somewhat irrational. Therefore the logical criterion alone is inadequate, as the following example is intended to show. In the edition of Dichtungen which Georg Heym's friends produced with a certain nonchalance in 1922 from his literary remains, a stanza of the poem "An das Meer" begins with the lines: "Und etwas tauchen aus der Flut, der matten, / Gesichter, wesenlos vom Totenreich." When

Carl Seelig produced the first complete edition of Heym's poems in 1947 he thought that the first line simply did not make sense, tried to reconstruct the graphical image and ended with the following conjecture: "Und Ätnas tauchen aus der Flut, der matten, / Gesichter, wesenlos vom Totenreich." Thus he had wrongly interpreted "etwas" as subject instead of adverbial; in this case the printed copy of 1922 is correct.[45]

If an editor assesses a passage only according to the criterion of sense, that is to say, only in terms of what he thinks possible for the author or for the text in question, there is always a danger that something unique, a neologism or an anomaly, may be removed from the text in favour of something more normal. A strange example occurs in the manuscripts of C. F. Meyer. The author had written: "Wie von grausem Bann befreit"; his sister made an alteration, presumably writing from dictation, and replaced the first two words by "Von dem", so that the passage now ran: "Von dem grausem Bann befreit". In print, it was rendered grammatically correct as: "Von dem grausen Bann befreit" (Poem no. 177 D5, l. 137). In the handbooks of historical German syntax and in the specialist literature on this subject there is no example from the nineteenth century of the strong form of the adjective after a definite article, as produced by Meyer's sister, so that it is easy to assume that the sister did not carry out the alteration to the end. But this anomaly occurs occasionally in Meyer's own uncorrected and unquestionably legible poetic manuscripts, for instance "Mit dem erschöpftem Pferd" (Poem no. 170 M6, l. 38); and again the reverse case which is perhaps even more noteworthy: "Mit beredten Mund" (Poem No. 158 M3, l. 44); again, in all the versions of the story "Angela Borgia" which appeared during Meyer's life-time: "Mit grausamen Genusse". It is significant that the manuscripts and the printed copy in the periodical still exhibit the strong form which one would expect at this point ("grausamem"), so that the supposedly faulty form does not derive from Meyer but from the compositor, that is to say, from another nineteenth-century user of the language, and Meyer did not alter it, but the editor of the prose in the historical-critical edition considered it to be a misprint and reversed it. C. F. Meyer was no linguistic innovator, and it is inconceivable to me that

he should have been alone in the nineteenth century in using inflexional forms in this way. If the forms are not recorded in the reference works and in other linguistic literature, this, and probably many similar instances, is caused by a vicious circle: the editors eliminate rare forms from the language of the edited text in the critical editions, because on the basis of their knowledge of the language, their ideas of the author, their reference books and the descriptions of historical grammar, metrics, etc., they consider them faulty. But these works of reference and their predecessors are based on just such editions purged by the editors.

In order to prevent this from happening, it is necessary to supplement the criterion of text-specific logic with the bibliographical criterion. This supplementation is necessary also because of the idiosyncrasy of the work of art referred to above, that it may violate the code which it has itself created. In the area of greatest authority, the autograph manuscript which is not simply a copy, one will be well-advised to admit no textual faults at all. The so-called slips of the pen, and such other disturbances of the writing process as may affect the author's projection of his work, occur, like all other deficiencies, against a specific meaningful background. A separation into deficiencies to be emended would only go to show that the editor's perspicuity is limited. One can of course object to the logical criterion by saying that the most insidious misprints are characterized as such, and remain undetected, because they do not generate a nonsensical text, but an impoverished, more trivial text, mutilated nevertheless. This is the case with the alteration of "Thurm und Chor" to "Thurm und Thor". Under quite specific circumstances, which have been described (Hawthorne's Blithedale Romance), such misprints would indeed need to be treated as textual faults. But in general one cannot discount the possibility that it is a matter of a "second generation" misprint which the author may not have recognised as a misprint, but which he registered and built into the semiotic system, that is to say, he authorised it and therefore it cannot later be reversed. If one regards cases of this sort as dubious, they are decided according to the editorial measures which we are recommending in favour of the textus receptus. This has the additional advantage that it is this text which is essential for the history of influence and reception.

A textual fault obliges the editor to intervene in the text. But emendation is only permitted in the edited text when the correction is unequivocal. In all other cases the textual fault or the gap in the text should be marked, that is to say, indicated by a mark in the text.

Suggested improvements should be discussed in the apparatus. In order to comply with the rigorous requirements of text establishment, we recommend that simple corruptions of the text and mistakes (suspected compositor's errors and slips of the pen), which the author may have overlooked and which do not meet the definition of the textual fault, should also be marked in this way in the text or in the margin. Appropriate to the problematical nature of our edited texts is also, to my mind, a minimal apparatus of emendation at the foot of the page in the text, corresponding to the markings in the text. Both these technical measures are intended to draw the attention of the reader to problems of textual crticism and to possible solutions while he is reading the text; apart from this they provide a reference to the main apparatus. In this critical apparatus all problematical passages should be discussed in detail, which has not been usual in German editions, and this discussion should be modelled on such exemplary "textual notes" as are found, for instance, in the editions of Fredson Bowers.

Notes