Textual criticism (or lower criticism) is a branch of literary criticism that is concerned with the identification and removal of transcription errors in the texts of manuscripts. Ancient scribes often made errors or alterations, while copying manuscripts by hand.1 Given a manuscript copy, several or many copies, but not the original document, the textual critic seeks to reconstruct the original text (the archetype or autograph) as closely as possible. The same processes can be used to attempt to reconstruct intermediate editions, or recensions, of a document's transcription history.2 The ultimate objective of the textual critic's work is the production of a "critical edition" containing a text most closely approximating the original.
There are three fundamental approaches to textual criticism: eclecticism, stemmatics, and copy-text editing. Techniques from the biological discipline of cladistics are currently also being used to determine the relationships between manuscripts.
Lower criticism is used to describe the contrast between textual criticism and "higher" criticism, which is the endeavor to establish the authorship, date, and place of composition of the original text. The origin of textual criticism is rooted in both the rise of modern historigraphy, which provided greater tools for textual analysis, and the work of religious scholars to answer the pressing questions about the origins of sacred texts.
Textual criticism has been practiced for over two thousand years. Early textual critics were concerned with preserving the works of antiquity, and this continued through the medieval period into early modern times until the invention of the printing press.
Many ancient works, such as the Bible and the Greek tragedies, survive in hundreds of copies, and the relationship of each copy to the original may be unclear. Textual scholars have debated for centuries which sources are most closely derived from the original, hence which readings in those sources are correct. Although biblical books that are letters, like Greek plays, presumably had one original, the question of whether some biblical books, like the gospels, ever had just one original has been discussed.3
In the English language, the works of Shakespeare have been a particularly fertile ground for textual criticism-both because the texts, as transmitted, contain a considerable amount of variation, and because the effort and expense of producing superior editions of his works have always been widely viewed as worthwhile.4 The principles of textual criticism, although originally developed and refined for works of antiquity, the Bible, and Shakespeare,5 have been applied to many works, extending backwards from the present to the earliest known written documents, in Mesopotamia and Ancient Egypt-a period of about five millennia.
The textual critic's ultimate objective is the production of a "critical edition." This contains a text most closely approximating the original, which is accompanied by an apparatus criticus (or critical apparatus) that presents:
- the evidence that the editor considered (names of manuscripts, or abbreviations called sigla),
- the editor's analysis of that evidence (sometimes a simple likelihood rating), and
- a record of rejected variants (often in order of preference).6
ProcessFolio from Papyrus 46, containing 2 Corinthians 11:33-12:9
Before mechanical printing, literature was copied by hand, and many variations were introduced by copyists. The age of printing made the scribal profession effectively redundant. Printed editions, while less susceptible to the proliferation of variations likely to arise during manual transmission, are nonetheless not immune to introducing variations from an author's autograph. Instead of a scribe miscopying his source, a compositor or a printing shop may read or typeset a work in a way that differs from the autograph.7 Since each scribe or printer commits different errors, reconstruction of the lost original is often aided by a selection of readings taken from many sources. An edited text that draws from multiple sources is said to be eclectic. In contrast to this approach, some textual critics prefer to identify the single best surviving text rather than combining readings from multiple sources.8
When comparing different documents, or "witnesses," of a single, original text, the observed differences are called variant readings, or simply variants or readings. It is not always apparent which single variant represents the author's original work. The process of textual criticism seeks to explain how each variant may have entered the text, either by accident (duplication or omission) or intention (harmonization or censorship), as scribes or supervisors transmitted the original author's text by copying it. The textual critic's task, therefore, is to sort through the variants, eliminating those most likely to be un-original, hence establishing a "critical text," or critical edition, that is intended to best approximate the original. At the same time, the critical text should document variant readings, so the relation of extant witnesses to the reconstructed original is apparent to a reader of the critical edition. In establishing the critical text, the textual critic considers both "external" evidence (the age, provenance, and affiliation of each witness) and "internal" or "physical" considerations (what the author and scribes, or printers, were likely to have done).3
The collation of all known variants of a text is referred to as a Variorum, namely a work of textual criticism whereby all variations and emendations are set side by side so that a reader can track how textual decisions have been made in the preparation of a text for publication.9 The Bible and the works of William Shakespeare have often been the subjects of variorum editions, although the same techniques have been applied with less frequency to many other works, such as Walt Whitman's Leaves of Grass.10 and the prose writings of Edward Fitzgerald.11
Eclecticism refers to the practice of consulting a wide diversity of witnesses to a particular original. The practice is based on the principal that the more independent two transmission histories are, the less likely they will be to reproduce the same errors. What one omits, the other may retain; what one adds, the other is unlikely to add. Eclecticism allows inferences to be drawn regarding the original text, based on the evidence of contrasts between witnesses.
Eclectic readings also normally give an impression of the number of witnesses to each available reading. Although a reading supported by the majority of witnesses is frequently preferred, this does not follow automatically. For example, a second edition of a Shakespeare play may include an addition alluding to an event known to have happened between the two editions. Although nearly all subsequent manuscripts may have included the addition, textual critics may reconstruct the original without the addition.
The result of the process is a text with readings drawn from many witnesses. It is not a copy of any particular manuscript, and may deviate from the majority of existing manuscripts. In a purely eclectic approach, no single witness is theoretically favored. Instead, the critic forms opinions about individual witnesses, relying on both external and internal evidence.12
Since the mid-nineteenth century, eclecticism, in which there is no a priori bias to a single manuscript, has been the dominant method of editing the Greek text of the New Testament (currently, the United Bible Society, 4th ed. and Nestle-Aland, 27th ed.). Even so, the oldest manuscripts of the Alexandrian text-type are the most favored, and the critical text has an Alexandrian disposition.13
External evidence is evidence of each physical witness, its date, source, and relationship to other known witnesses. Critics will often prefer the readings supported by the oldest witnesses. Since errors tend to accumulate, older manuscripts should have fewer errors. Readings supported by a majority of witnesses are also usually preferred, since these are less likely to reflect accidents or individual biases. For the same reasons, the most geographically diverse witnesses are preferred. Some manuscripts show evidence that particular care was taken in their composition, for example, by including alternative readings in their margins, demonstrating that more than one prior copy (exemplar) was consulted in producing the current one. Other factors being equal, these are the best witnesses.
There are many other more sophisticated considerations. For example, readings that depart from the known practice of a scribe or a given period may be deemed more reliable, since a scribe is unlikely on his own initiative to have departed from the usual practice.14
Internal evidence is evidence that comes from the text itself, independent of the physical characteristics of the document. Various considerations can be used to decide which reading is the most likely to be original. Sometimes these considerations can be in conflict.14
Two common considerations have Latin names lectio brevior (shorter reading) and lectio difficilior (more difficult reading). The first is the general observation that scribes tended to add words, for clarification or out of habit, more often than they removed them. The second, lectio difficilior potior (the harder reading is stronger), recognizes the tendency for harmonization-resolving apparent inconsistencies in the text. Applying this principle leads to taking the more difficult (unharmonized) reading as the more likely to be the original. Such cases also include scribes simplifying and smoothing texts they did not fully understand.15 Some developing research, however, suggests that this principle cannot be applied universally, as is the case with the Book of Revelation where much of the text is difficult to understand and interpret.
Another scribal tendency is called homoioteleuton, meaning "same endings." Homoioteleuton occurs when two words/phrases/lines end with the same sequence of letters. The scribe, having finished copying the first, skips to the second, omitting all intervening words. Homeoarchy refers to eye-skip when the beginnings of two lines are similar.
The critic may also examine the other writings of the author to decide what words and grammatical constructions match his style. The evaluation of internal evidence also provides the critic with information that helps him evaluate the reliability of individual manuscripts. Thus, the consideration of internal and external evidence is related.
After considering all relevant factors, the textual critic seeks the reading that best explains how the other readings would arise. That reading is then the most likely candidate to have been original.
Canons of textual criticismLuke 11:2 in Codex Sinaiticus
Various scholars have developed guidelines, or canons of textual criticism, to guide the exercise of the critic's judgment in determining the best readings of a text. One of the earliest was Johann Albrecht Bengel (1687-1752), who in 1734 produced an edition of the Greek New Testament. In his commentary, he established the rule Proclivi scriptioni praestat ardua, ("the harder reading is to be preferred") 16
Johann Jakob Griesbach (1745-1812) published several editions of the New Testament. In his 1796 edition Novum Testamentum Graece 17, he established 15 critical rules. Among them was a variant of Bengel's rule, Lectio difficilior potior, "the hardest reading is best." Another was Lectio brevior praeferenda, "the shorter reading is best," based on the idea that scribes were more likely to add than to delete.18 This rule cannot be applied uncritically, as scribes may omit material inadvertently.
Brooke Foss Westcott (1825-1901) and Fenton J. A. Hort (1828-1892) published an edition of the New Testament in 1881. They proposed nine critical rules, including a version of Bengel's rule,
"The reading is less likely to be original that shows a disposition to smooth away difficulties."
They also argued that "Readings are approved or rejected by reason of the quality, and not the number, of their supporting witnesses," and that "The reading is to be preferred that most fitly explains the existence of the others."19
Many of these rules, although originally developed for Biblical textual criticism, have wide applicability to any text susceptible to errors of transmission.
Limitations of eclecticism
Since the canons of criticism are highly susceptible to interpretation, and at times even contradict each other, they can often be employed to justify any result that fits the text critic's aesthetic or theological agenda. Starting in the nineteenth century, scholars sought more rigorous methods to guide editorial judgment. Best-text editing (a complete rejection of eclecticism) emerged as one extreme. Stemmatics and copy-text editing-while both eclectic, in that they permit the editor to select readings from multiple sources-sought to reduce subjectivity by establishing one or a few witnesses, presumably as more "objective" criteria.
Stemmatics or stemmatology is a rigorous approach to textual criticism. Karl Lachmann (1793-1851) greatly contributed to making this method famous, even though he did not invent it (see Timpanaro, The genesis of Lachmann's method). The method takes its name from the stemma, "family tree," which shows the relationships of the surviving witnesses. The family tree is also referred to as a cladorama.20 The method works from the principle that a "community of error implies community of origin." That is, if two witnesses have a number of errors in common, it may be presumed that they were derived from a common intermediate source, called a hyparchetype. Relations between the lost intermediates are determined by the same process, placing all extant manuscripts in a family tree or stemma codicum descended from a single archetype. The process of constructing the stemma is called recension, or the Latin recensio.21
Having completed the stemma, the critic proceeds to the next step, called selection or selectio, where the text of the archetype is determined by examining variants from the closest hyparchetypes to the archetype and selecting the best ones. If one reading occurs more often than another at the same level of the tree, then the dominant reading is selected. If two competing readings occur equally often, then the editor uses his judgment to select the correct reading.22
After selectio, the text may still contain errors, since there may be passages where no source preserves the correct reading. The step of examination, or examinatio is applied to find corruptions. Where the editor concludes that the text is corrupt, it is corrected by a process called "emendation," or emendatio (also sometimes called divinatio). Emendations not supported by any known source are sometimes called conjectural emendations.23
The process of selectio resembles eclectic textual criticism, but applied to a restricted set of hypothetical hyparchetypes. The steps of examinatio and emendatio resemble copy-text editing. In fact, the other techniques can be seen as special cases of stemmatics, but in which a rigorous family history of the text cannot be determined but only approximated. If it seems that one manuscript is by far the best text, then copy text editing is appropriate, and if it seems that a group of manuscripts are good, then eclecticism on that group would be proper.
The Hodges-Farstad edition of the Greek New Testament attempts to use stemmatics for some portions.
Limitations and criticism
The stemmatic method assumes that each witness is derived from one, and only one, predecessor. If a scribe refers to more than one source when creating his copy, then the new copy will not clearly fall into a single branch of the family tree. In the stemmatic method, a manuscript that is derived from more than one source is said to be contaminated.
The method also assumes that scribes only make new errors; they do not attempt to correct the errors of their predecessors. When a text has been improved by the scribe, it is said to be sophisticated, but "sophistication" impairs the method by obscuring a document's relationship to other witnesses, and making it more difficult to place the manuscript correctly in the stemma.
The stemmatic method requires the textual critic to group manuscripts by commonality of error. It is required, therefore, that the critic can distinguish erroneous readings from correct ones. This assumption has often come under attack. W. W. Greg noted, "That if a scribe makes a mistake he will inevitably produce nonsense is the tacit and wholly unwarranted assumption."24
The critic Joseph Bédier (1864-1938) launched a particularly withering attack on stemmatics in 1928. He surveyed editions of medieval French texts that were produced with the stemmatic method, and found that textual critics tended overwhelmingly to produce trees divided into just two branches. He concluded that this outcome was unlikely to have occurred by chance, and that therefore, the method was tending to produce bipartite stemmas regardless of the actual history of the witnesses. He suspected that editors tended to favor trees with two branches, as this would maximize the opportunities for editorial judgment (as there would be no third branch to "break the tie" whenever the witnesses disagreed). He also noted that, for many works, more than one reasonable stemma could be postulated, suggesting that the method was not as rigorous or as scientific as its proponents had claimed.
The stemmatic method's final step is emendatio, also sometimes referred to as "conjectural emendation." But in fact, the critic employs conjecture at every step of the process. Some of the method's rules that are designed to reduce the exercise of editorial judgment do not necessarily produce the correct result. For example, where there are more than two witnesses at the same level of the tree, normally the critic will select the dominant reading. However, it may be no more than fortuitous that more witnesses have survived that present a particular reading. A plausible reading that occurs less often may, nevertheless, be the correct one.25
Lastly, the stemmatic method assumes that every extant witness is derived, however remotely, from a single source. It does not account for the possibility that the original author may have revised his work, and that the text could have existed at different times in more than one authoritative version.
Copy-text editingPage from Codex Vaticanus Graece 1209, B/03
When copy-text editing, the scholar fixes errors in a base text, often with the help of other witnesses. Often, the base text is selected from the oldest manuscript of the text, but in the early days of printing, the copy text was often a manuscript that was at hand.
Using the copy-text method, the critic examines the base text and makes corrections (called emendations) in places where the base text appears wrong to the critic. This can be done by looking for places in the base text that do not make sense or by looking at the text of other witnesses for a superior reading. Close-call decisions are usually resolved in favor of the copy-text.
The first published, printed edition of the Greek New Testament was produced by this method. Erasmus (1466 - 1536), the editor, selected a manuscript from the local Dominican monastery in Basle and corrected its obvious errors by consulting other local manuscripts. The Westcott and Hort text, which was the basis for the Revised Version of the English Bible, also used the copy-text method, using the Codex Vaticanus as the base manuscript.
McKerrow's concept of copy-text
The bibliographer Ronald B. McKerrow introduced the term copy-text in his 1904 edition of the works of Thomas Nashe, defining it as "the text used in each particular case as the basis of mine." McKerrow was aware of the limitations of the stemmatic method, and believed it was more prudent to choose one particular text that was thought to be particularly reliable, and then to emend it only where the text was obviously corrupt. The French critic Joseph Bédier likewise became disenchanted with the stemmatic method, and concluded that the editor should choose the best available text, and emend it as little as possible.
In McKerrow's method as originally introduced, the copy-text was not necessarily the earliest text. In some cases, McKerrow would choose a later witness, noting that "if an editor has reason to suppose that a certain text embodies later corrections than any other, and at the same time has no ground for disbelieving that these corrections, or some of them at least, are the work of the author, he has no choice but to make that text the basis of his reprint."26
By 1939, in his Prolegomena for the Oxford Shakespeare, McKerrow had changed his mind about this approach, as he feared that a later edition-even if it contained authorial corrections-would "deviate more widely than the earliest print from the author's original manuscript." He therefore concluded that the correct procedure would be "produced by using the earliest 'good' print as copy-text and inserting into it, from the first edition which contains them, such corrections as appear to us to be derived from the author." But, fearing the arbitrary exercise of editorial judgment, McKerrow stated that, having concluded that a later edition had substantive revisions attributable to the author, "we must accept all the alterations of that edition, saving any which seem obvious blunders or misprints."27
W. W. Greg's rationale of copy-text
Anglo-American textual criticism in the last half of the twentieth century came to be dominated by a landmark 1950 essay by Sir Walter W. Greg, "The Rationale of Copy-Text." Greg proposed:
A distinction between the significant, or as I shall call them 'substantive', readings of the text, those namely that affect the author's meaning or the essence of his expression, and others, such in general as spelling, punctuation, word-division, and the like, affecting mainly its formal presentation, which may be regarded as the accidents, or as I shall call them 'accidentals', of the text.28
Greg observed that compositors at printing shops tended to follow the "substantive" readings of their copy faithfully, except when they deviated unintentionally; but that "as regards accidentals they will normally follow their own habits or inclination, though they may, for various reasons and to varying degrees, be influenced by their copy."29
The true theory is, I contend, that the copy-text should govern (generally) in the matter of accidentals, but that the choice between substantive readings belongs to the general theory of textual criticism and lies altogether beyond the narrow principle of the copy-text. Thus it may happen that in a critical edition the text rightly chosen as copy may not by any means be the one that supplies most substantive readings in cases of variation. The failure to make this distinction and to apply this principle has naturally led to too close and too general a reliance upon the text chosen as basis for an edition, and there has arisen what may be called the tyranny of the copy-text, a tyranny that has, in my opinion, vitiated much of the best editorial work of the past generation.30
Greg's view, in short, was that the "copy-text can be allowed no over-riding or even preponderant authority so far as substantive readings are concerned." The choice between reasonable competing readings, he said:
Will be determined partly by the opinion the editor may form respecting the nature of the copy from which each substantive edition was printed, which is a matter of external authority; partly by the intrinsic authority of the several texts as judged by the relative frequency of manifest errors therein; and partly by the editor's judgment of the intrinsic claims of individual readings to originality-in other words their intrinsic merit, so long as by 'merit' we mean the likelihood of their being what the author wrote rather than their appeal to the individual taste of the editor.31
Although Greg argued that an editor should be free to use his judgment to choose between competing substantive readings, he suggested that an editor should defer to the copy-text when "the claims of two readings… appear to be exactly balanced.… In such a case, while there can be no logical reason for giving preference to the copy-text, in practice, if there is no reason for altering its reading, the obvious thing seems to be to let it stand."32 The "exactly balanced" variants are said to be indifferent.
Editors who follow Greg's rationale produce eclectic editions, in that the authority for the "accidentals" is derived from one particular source (usually the earliest one) that the editor considers to be authoritative, but the authority for the "substantives" is determined in each individual case according to the editor's judgment. The resulting text, except for the accidentals, is constructed without relying predominantly on any one witness.
W. W. Greg did not live long enough to apply his rationale of copy-text to any actual editions of works. His rationale was adopted and significantly expanded by Fredson Bowers (1905-1991). Starting in the 1970s, G. Thomas Tanselle (1934-) vigorously took up the method's defense and added significant contributions of his own. Greg's rationale as practiced by Bowers and Tanselle has come to be known as the "Greg-Bowers" or the "Greg-Bowers-Tanselle" method.
Application to works of all periodsWilliam Shakespeare, A Midsummer Night's Dream
In his 1964 essay, "Some Principles for Scholarly Editions of Nineteenth-Century American Authors," Bowers said that "the theory of copy-text proposed by Sir Walter Greg rules supreme".33 Bowers's assertion of "supremacy" was in contrast to Greg's more modest claim that "My desire is rather to provoke discussion than to lay down the law".34
Whereas Greg had limited his illustrative examples to English Renaissance drama, where his expertise lay, Bowers argued that the rationale was "the most workable editorial principle yet contrived to produce a critical text that is authoritative in the maximum of its details whether the author be Shakespeare, Dryden, Fielding, Nathaniel Hawthorne, or Stephen Crane. The principle is sound without regard for the literary period."35 For works where an author's manuscript survived - a case Greg had not considered - Bowers concluded that the manuscript should generally serve as copy-text. Citing the example of Nathaniel Hawthorne, he noted:
When an author's manuscript is preserved, this has paramount authority, of course. Yet the fallacy is still maintained that since the first edition was proofread by the author, it must represent his final intentions and hence should be chosen as copy-text. Practical experience shows the contrary. When one collates the manuscript of The House of the Seven Gables against the first printed edition, one finds an average of ten to fifteen differences per page between the manuscript and the print, many of them consistent alterations from the manuscript system of punctuation, capitalization, spelling, and word-division. It would be ridiculous to argue that Hawthorne made approximately three to four thousand small changes in proof, and then wrote the manuscript of The Blithedale Romance according to the same system as the manuscript of the Seven Gables, a system that he had rejected in proof.36
Following Greg, the editor would then replace any of the manuscript readings with substantives from printed editions that could be reliably attributed to the author: "Obviously, an editor cannot simply reprint the manuscript, and he must substitute for its readings any words that he believes Hawthorne changed in proof.37
Uninfluenced final authorial intention
McKerrow had articulated textual criticism's goal in terms of "our ideal of an author's fair copy of his work in its final state."38 Bowers asserted that editions founded on Greg's method would "represent the nearest approximation in every respect of the author's final intentions."39 Bowers stated similarly that the editor's task is to "approximate as nearly as possible an inferential authorial fair copy."40 Tanselle notes that, "Textual criticism… has generally been undertaken with a view to reconstructing, as accurately as possible, the text finally intended by the author".41
Bowers and Tanselle argue for rejecting textual variants that an author inserted at the suggestion of others. Bowers said that his edition of Stephen Crane's first novel, Maggie, presented "the author's final and uninfluenced artistic intentions."42 In his writings, Tanselle refers to "unconstrained authorial intention" or "an author's uninfluenced intentions."43 This marks a departure from Greg, who had merely suggested that the editor inquire whether a later reading "is one that the author can reasonably be supposed to have substituted for the former",44 not implying any further inquiry as to why the author had made the change.
Tanselle discusses the example of Herman Melville's Typee. After the novel's initial publication, Melville's publisher asked him to soften the novel's criticisms of missionaries in the South Seas. Although Melville pronounced the changes an improvement, Tansell