Some Notes on Genealogical Methods in Textual Criticism

Some Notes on Genealogical Methods in Textual CriticismAuthor(s): Vinton A. DearingSource: Novum Testamentum, Vol. 9, Fasc. 4 (Oct., 1967), pp. 278-297Published by: BRILLStable URL: http://www.jstor.org/stable/1560358 .

Accessed: 29/12/2013 18:18

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

BRILL is collaborating with JSTOR to digitize, preserve and extend access to Novum Testamentum.

http://www.jstor.org

This content downloaded from 66.77.17.54 on Sun, 29 Dec 2013 18:18:32 PMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=bap

http://www.jstor.org/stable/1560358?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


SOME NOTES ON GENEALOGICAL METHODS IN TEXTUAL CRITICISM

BY

VINTON A. DEARING London

In the outstandingly full and lucid exposition of textual criticism in The Text of the New Testament, BRUCE M. METZGER lists as the third type of internal evidence "to be taken into account in

evaluating variant readings" what he describes as "the genealogical relationships of texts and families of witnesses." 1) Most textual critics of the Bible, however, give comparatively little attention to textual genealogy, accepting the word of a few whose methods are capable of better formulation and application-with perhaps no corresponding revision of former conclusions, but no one can say until improved methods are put to use. This paper will describe more fully two improved methods for determining "the genealogical relationships of texts," for Professor METZGER, who mentions them, does not explain how to carry them out. One method developed since Professor METZGER'S book will also be described. The three methods supplement each other, forming together a single and consistent method for solving problems of

genealogical relationships. Professor METZGER begins his chapter on "Modern Methods of

Textual Criticism" with a clear and concise account of what he calls "The Classical Method of Textual Criticism," or "the genealogical method" (pp. I56-59; italics mine). The simple rules of this method will be familiar to many readers, and in any case will serve as an introduction to the more difficult concepts to be set out later. The central passage of Professor METZGER'S description of the method is as follows:

The basic principle which underlies the process of constructing a stemma, or family tree, of manuscripts is that, apart from accident, identity of reading implies identity of origin. By way of

1) The Text of the New Testament: Its Transmission, Corruption, and Restoration (New York, 1964), p. 209.



NOTES ON GENEALOGICAL METHODS IN TEXTUAL CRITICISM 279

example, suppose that there are seven manuscripts of an ancient book, and that ... one of them (which we may designate A) stands

apart, showing no great similarity to any of the other six, while B, C, and D, on the one hand, and E, F, and G, on the other hand, greatly resemble each other, though differing somewhat from the rest. We can express this by saying that B, C, and D form a family, descended from a hypothetical common ancestor which we may call X, and that E, F, and G form another family, descended from a hypothetical ancestor which we may call Y. ... It is possible to ... deduce ... a still more remote ancestor which we may call Z, the hypothetical archetype of all the manuscripts. Thus the pedigree of all ten manuscripts (the seven extant and the three hypothetical) would be as follows:

z

I I x Y

A B C D E F G

This method is essentially vitiated by the fact that given the

principle appealed to, and the evidence adduced by Professor METZGER, X or Y could equally well be the archetype, or the

archetype could be at any of a number of other places in the

linkage between the manuscripts, as was demonstrated by Sir WALTER GREG in The Calculus of Variants 1). GREG gives the full

1) Professor METZGER gives an account of GREG'S work (pp. 165-166), but not in enough detail to bring this out. Like a number of other critics, he objects to what he calls "a needless proliferation of pseudo-mathematical symbols" in GREG'S Calculus, but GREG'S notational system can be justified. It is not "pseudo-mathematical" unless the symbols in modern logic are so, and it performs the same function, that is, it allows certain fundamental qualities of the matter under consideration to be examined by themselves. In this instance, the problem is what can be made of the configurations of agreement and disagreement among manuscripts as distinct from the values of the readings. GREG may not have needed his symbols, and manipulating them may not have showed him the way to his conclusions, but the fact remains that he was the first textual critic to recognize that configurations of agreement and disagreement do not identify ancestors and archetypes, and that he was able to use his symbols to prove his point. My summary farther on in this paper of the rest of GREG'S work avoids his notation to avoid objections like Professor METZGER'S, but in extending GREG'S rules I have often wished I had extended his notation as well, since truth and implication are so much more easily discoverable and demonstrable in this way. The notation used in lattice algebra was devised for geometric dia-



VINTON A. DEARING

set of alternate possibilities with six manuscripts. Below are a few of the additional possibilities with Professor METZGER'S seven:

w

x Y x

z z z z Y Xx x y y

~ '-I- I II-II '-A- --, r-? - - I--h ! ? A E F G BCD A BCD EFG A B C DE F G AE FG CD B

The method Professor METZGER describes produces a single stemma because it is not in fact based solely on the principle that, apart from accident, identity of reading implies identity of origin. Instead, there is in this particular genealogical method an un-

expressed assumption that the immediate hypothetical ancestor of one group of manuscripts will not be a more remote ancestor of another group of manuscripts, an assumption that, once it is

recognized, will be seen to have no logical basis. Professor METZGER closes his account of this method by saying

that "To the extent that the manuscripts have a 'mixed' ancestry 1) the genealogical relations among them become progressively more

complex and obscure to the investigator" who uses it. This is, of

course, a crippling limitation as far as Biblical scholars are concerned, even if the method were otherwise satisfactory.

The method or assemblage of methods now to be described has neither the logical flaw nor the operational limitation of that described by Professor METZGER. It consists of two steps. The first

step is the determination of what may be termed the essential

linkage between the manuscripts. The essential linkage in the

foregoing example is the connection of the extant manuscripts through X, Y, and Z, which is essentially the same wherever the

archetype may prove to be, e.g., the lines between B, C, or D and

any of the other four manuscripts always pass through X. The second step is the determination of the archetype. Each step has its own principle and its own rules for the application of its

grams, of which stemmas are a sub-class, and should prove very useful in formal proofs of genealogical methods.

1) ,,That is, when a copyist has had two or more manuscripts before him and has followed sometimes one, sometimes the other; or, as sometimes happened, when a scribe copied a manuscript from one exemplar and corrected it against another." Both quotations from p. I59.

280




principle. Unfortunately, the justification of the method cannot be

adequately demonstrated in advance, but those readers who wish a glimpse of the goal may perhaps obtain it in and between the italicized passages on pp. 295 and 296.

I.-PRINCIPLES

The principle that brings about the two-fold division of the method, namely, that a stemma is not changed in its essential

linkage by taking any point along its branches, including the extremities, as the topmost point, has just been illustrated. The same principle requires that the determination of the essential

linkage precede the determination of the archetype. The principle for determining the essential linkage between the

manuscripts is not sufficiently clearly expressed by saying that

"apart from accident, identity of reading implies identity of

origin." Strictly interpreted, this phrasing means that no extant

manuscript can be a copy or stand at the end of a chain of copies of

any other, a corollary that is not, I believe, a part of any genealogical method, though the similar corollary that no extant manu-

script can be the archetype is tacit in most. Furthermore, accident is not the only reason why identity of reading does not always imply identity of origin even in the loose sense: a manuscript may agree with another against its own exemplar because it has been corrected by memory or direct consultation of the other manuscript or has been conjecturally emended, or because some characteristic of the text has led the scribes to make the same mistake. Correcting a manuscript by direct consultation of a manuscript other than its

exemplar is only a minor instance of copying a manuscript from two exemplars. The major instances can be dealt with like the minor if we introduce the idea of a principal exemplar for each

copy to which alone the "identity of reading" principle will apply. And "accident" itself covers at least two possibilities: mere coin- cidence, and unintended irruption of a memory in the second scibe of the text of the other manuscript. If we give to all these

possibilities the general term "abnormal agreements," we can

express the principle more adequately as follows: abnormalities

apart, two manuscripts will agree only if one has been copied from the other or stands at the end of a chain of copies of the other, or if both have been copied from a third or stand at the ends of chains of copies of the third.



VINTON A. DEARING

This principle is self-evident-it reflects the purpose and

necessary process of copying one manuscript from another-and as will be seen it allows the assignment of hypothetical intermediaries (this is a better term than hypothetical ancestors, because, as the last stemma above shows, X is not necessarily the ancestor of B, C, and D), and the reconstruction of their texts.

The principle for determining the archetype is inherent in the definition of a stemma. In a properly drawn stemma, all the

"good" readings in the manuscripts can be traced back to the

archetype, since the archetype is by definition the earliest identifiable form of the text before emendation. The distribution of the more original readings therefore determines the form of the stemma.

II.-RULES

A-First Step: Determining the Essential Linkage

The rules for this step will be discussed in three groups, the rules in each group being in order of application, but the groups in reverse order of application. The latter arrangement is dictated by the fact that the rules in the first group are easiest to understand, and that once understood they facilitate the understanding of the rules in the second group, and so on. This order of the groups happens also to be the order of their discovery in the history of textual criticism. In each group, the explanation of the rules is followed by directions for applying them fully and consistently.

The truth of the rules will be more obvious if the implications of their underlying principle are first clearly understood, namely that, abnormalities apart, if two manuscripts agree against a third, the third cannot be part of any chain of copies linking the other two to each other or to a common ancestor, and that if the third manu-

script agrees with a fourth when the first two agree against it, the fourth manuscript stands in the same relationship to the first two as the third while, conversely, the first two stand in the same

relationship to the third and fourth. Since the proviso "abnormalities apart" is normally impossible in real situations, the third

group of rules consists of directions for avoiding it; the rules in the first two groups are easiest to understand if the proviso is allowed to stand for the time being.

I.-Simple Variations

Greg laid down the three rules for determining the essential

282




linkage in cases where the manuscripts exhibit only two alternate

readings, or what he called "simple variations". The first rule is that when all but one of the manuscripts agree, which he called a

"type-i" variation, the one manuscript must be terminal in the

linkage. Thus, using a colon to separate the sigla according to the

agreements of the manuscripts, the occurence of type-I variations A: BCDEFG indicates that manuscript A will stand at an extremity of the linkage.

GREG'S second rule is that when two or more manuscripts agree against two or more others, which he called a "type-2" variation, each group of manuscripts must have its own common inter-

mediary. Thus, the occurrence of type-2 variations ABCD: EFG indicates that ABCD will have a common intermediary different from the common intermediary for EFG.

GREG'S third rule is that the common intermediary in a group of manuscripts will be hypothetical only if all the manuscripts in the group are terminal. If one is not terminal, it will be the common

intermediary for its group; if two are not terminal, they are

duplicates and either manuscript may be reckoned the common

intermediary; if only one is terminal, the other, or any one of the others, since if there are more than one they are duplicates, is the common intermediary for the group, even though the appelation "common" is not strictly applicable in this circumstance; if none are terminal, all are duplicates, the equivalent of a single terminal

manuscript, so neither "common" nor "intermediary" is strictly applicable.

To insure that GREG'S rules are uniformly applied it is wise to

proceed in the following three-step routine. First, identify all the terminal manuscripts. Second, systematically examine all the

groups of half or fewer of the manuscripts in type-2 variations, starting with the smallest groups. For each such group assign a

hypothetical intermediary if all its manuscripts are terminal, and otherwise assign one of its manuscripts as the intermediary or treat the group as a terminal manuscript, according to GREG'S

third rule-unless intermediaries have already been assigned to some or all of the manuscripts; in the latter case, treat the intermediaries the same as terminal manuscripts and ignore the manuscripts and any intermediaries to which these intermediaries have been assigned. Third, if more than two manuscripts or intermediaries remain that have not been assigned an intermediary,



VINTON A. DEARING

assign them an intermediary; if there are only two, connect them

directly. The foregoing stemmas all imply that if the agreements among

the manuscripts are not abnormal and occur in simple variations

only, then there must be examples of all seven of the possible type-I variations, and of only two of the possible type-2 variations, ABCD: EFG and AEFG: BCD. With all seven possible type-I variations, making all seven manuscripts terminal, the type-2 variation ABCD: EFG requires the assignment of a hypothetical intermediary to the group EFG and the type-2 variation AEFG: BCD requires the assignment of a hypothetical intermediary to the

group BCD (if B: ACDEFG had not occurred, B would be the

intermediary, and so on). There are no other groups of half the

manuscripts or less. The terminal manuscript A and the two intermediaries have not been assigned an intermediary. This

requires the assignment of a hypothetical intermediary for the three

(if A: BCDEFG had not occurred, the two intermediaries would be connected to A; if there were no manuscript A, the two intermediaries would be connected to each other).

GREG'S rules result in assigning a minimum of intermediaries, a very important matter when it comes to interpreting the resulting stemmas as in fact patterns of descent. From the practical point of view, that is what they are, but it is possible, for example, for two of three manuscripts with a single common hypothetical ancestor to have had in fact a common ancestor between them and the

hypothetical ancestor. In that event, of course, the ancestor of the two was a perfect copy of the ancestor of all three, that is, a

duplicate, not requiring a separate appearance in the stemma any more than a photostat of one of the extant manuscripts. GREG takes the practical point of view, remarking that a possible cause without a discernible effect can be ignored.

2.-Complex Variations

Of course, not all variations are necessarily "simple"; some or all can be "complex," that is, cases in which the manuscripts exhibit more than two alternate readings 1). Complex variations

1) In this connection it should be noted that if one manuscript omits a word and another omits this and the following word the two manuscript do not agree in omitting the first word, and similarly that if two manuscripts omit a pair of words, while two other manuscripts disagree as to the first

284




are not amenable to GREG'S rules, but most can be made amenable to them by three more rules 1), and most of the rest will be found to fit the linkage thus established, as A: BD: C: EFG will fit the

linkage in Professor METZGER'S example. If A and B agree against C when all three disagree with the

other manuscripts, as in AB: C: D: EFG, then C cannot be an

intermediary in any chain or chains of copies that link A and B, that is, C may not lie along a line A-B. If A and D also agree against C when B and D do not exhibit abnormal agreements as in AD: B: C: EFG, then C cannot be an intermediary in any chain or chains of copies that link A and D, that is, C may not lie along a line A-D. It follows that C cannot be intermediary in any chain or chains of copies that link A, B, and D, that is, C may not lie

along a line D-A-B. (When the linkage is determined, the line

segments may prove partly to coincide, diverging from a hypothetical intermediary instead of from A, but the principle is the

same). The overlap of the group AB with the group AD thus makes a synthetic group ABD. When such an overlap, or a series of them, makes a synthetic group ABDEFG, C will be necessarily terminal in the linkage, just as if the six had agreed against it in a type-I variation. This situation may then be recorded by writing a

synthetic type-I variation to be used in constructing the linkage. Thus the synthetic variation C: ABDEFG would be written from

word of the pair, the two manuscripts omitting the words do not agree in the matter of the first word. A "non-reading" cannot be subdivided. This was first pointed out by HILL, but it is too small a detail to find a place in Professor METZGER'S account of HILL'S work (pp. I66-I67). Professor METZGER objects to HILL'S statement that a stemma can show a medieval manuscript derived from a modern one, but what HILL has in mind is that if the modern manuscript is apparently an exact copy of an older one now lost, it may replace the older one in the stemma; thus one of Sir ROBERT COTTON'S Chaucer manuscripts, which was destroyed by fire, is represented in its stemma by an eighteenth-century transcript, although, since the shelf- mark of the oiiginal is known, it can be used in the stemma instead of the shelf-mark of the copy.

1) The rules that follow are re-expressions of points made in my Manual of Textual Analysis of which Professor METZGER has been kind enough to give an account (pp. I67-I69), but one too brief to mention them or the method there proposed for eliminating the proviso "abnormalities apart," a method superseded by that given in the next subsection of this paper. It is not within the province of this paper to defend matters in my work that Professor METZGER takes exception to but that have no relevance to problems of the descent of manuscripts. Further on, I shall have something to say in defense of computers.



VINTON A. DEARING

such variations as AB: C: DEFG and AC: BDEF: G, or AB: C: DE: FG, ABD: C: E: FG, and ABC: D: EFG.

The first rule supplementary to GREG'S is therefore that when a

synthetic group of manuscripts includes all but one, that one must be terminal.

Similarly, if A and B agree against C and D when C and D agree and when all the other manuscripts disagree with these four, as in AB: CD: E: FG, then neither C, D, nor any hypothetical manu-

script agreeing with them can be an intermediary in any chain or chains of copies that link A and B, and neither A, B, nor any hypothetical manuscript agreeing with them can be an inter-

mediary in any chain or chains of copies that link C and D; that is, no point on the line C-D can lie along the line A-B. If A and E also agree against C and D under normal circumstances, as in AE: B: CD: FG, then no point on the line C-D can lie along the line A-E. It follows that no point along the line C-D can lie along the line E-A-B. (If, however, the variations were only AB: CD: E: FG and AE: B: C: D: FG, a point on the line C-D could lie

along the line E-A-B, i.e., along such part of the E-A segment as does not coincide with the A-B segment. As before, the exact

relationships of the segments will not appear until the linkage is

determined.) The overlap of the group AB with the group AE makes a

synthetic group ABE. If such an overlap or a series of them can make a synthetic group ABEFG from variations in which C and D

agree, or, if they disagree, exactly one agrees with at least one other

manuscript, then ABEFG must have a common intermediary different from CD just as if the five had agreed against the two in a

type-2 variation. This situation may then be recorded by writing a

synthetic type-2 variation to be used in constructing the linkage. Thus the synthetic variation AB: CDEFG would be written from such variations as AB:CD:EFG and ABC:DE:FG. When a group, whether original or synthetic, includes one or more simple groups, whether original or synthetic, the group may be divided without external agreement provided that only simple groups are divided and exactly one subdivision of each group so divided agrees with at least one manuscript outside the simple group. Thus, if AB is a type-2 group, then A:BC:DEF:G and A:BCDE:FG allow the

writing of the synthetic variation ABC:DEFG. The second rule supplementary to Greg's is therefore that when

286




an original or a synthetic group of manuscripts occurring in variations that make a synthetic group of all the other manuscripts is always found undivided, or divided so that exactly one subdivision agrees with at least one manuscript outside the group, or divided so that only type-2 subgroups are split and exactly one subdivision of each of these agrees with at least one manuscript outside the subgroup, then each of the two groups must have its own common intermediary.

If the simple variations, original or synthetic, require that only two manuscripts be terminal (counting any groups of duplicate manuscripts therein as single terminal manuscripts according to Greg's third rule), then one more set of synthetic type-2 variations may be obtainable. Any complex group with only one terminal manuscript as just defined can be written as a group in a synthetic simple variation, because subsequent application of Greg's rules will never require assigning a hypothetical intermediary to any group. If there are more than two terminal manuscripts as just defined, then making synthetic simple variations from the complex, groups that have only one terminal manuscript and applying GREG'S rules will lead to the assignment of hypothetical intermediaries, and in every such case it will be found, upon testing the results against the basic principle, that the hypothetical intermediary is unnecessary and that any of the manuscripts to which the hypothetical intermediary was assigned can be intermediary between any of the others.

To modify our former example, if the only simple variations were A: BCDEFG, making A terminal, and ABCD: EFG, leaving E, F, and G duplicates and the equivalent of a single terminal manuscript, then the occurrence of the complex variation AB: C: DEFG would indicate that B was the intermediary for the group AB and that D was the intermediary for the group D(EFG), leaving C as the intermediary between B and D.

The third rule supplementary to GREG'S is, therefore, that when there are only two terminal manuscripts (counting groups of duplicate manuscripts as single terminal manuscripts according to GREG'S third rule), then if any group in a complex variation includes only one of these terminal manuscripts the remaining manuscripts must be treated as a synthetic group and each group must have its own common intermediary. (It will be found that in the resulting linkage the manuscripts most closely related will be



VINTON A. DEARING

most alike, as required by the principle underlying the rules.)

To insure that the three rules supplementary to GREG'S are

uniformly applied, it is wise to proceed in the following three-step routine. First, systematically make all possible overlaps of the

groups in the variations, simple as well as complex, and of the

resulting synthetic groups with the other original groups and with each other. Write synthetic type-I variations whenever a synthetic group includes all but one manuscript. Mark any synthetic group that does not exist also as an original group and cannot be made

up without using original groups that are the only true groups in their variations (i.e., can only be made up by using at least one such variation as A:B:CDEFG). Second, starting with the smallest

group or groups systematically match each group in each complex variation against each of the unmarked synthetic groups, and each of the unmarked synthetic groups against every other. If any two of these groups include all the manuscripts without sharing any, reexamine the complex variations to see if every manuscript in one

group agrees at least once with at least one other in the same group against all the manuscripts in the other group, when these agree, or, if these disagree, when exactly one of the subdivisions agrees with at least one manuscript outside the group or the subdivisions

split only type-2 subgroups and exactly one subdivision of each of these agrees with at least one manuscript outside the subgroup. Write synthetic type-2 variations whenever the required conditions are met. Third, systematically examine the simple variations, original and synthetic, to see if there are only two terminal manus-

cripts, counting groups of duplicate manuscripts as single terminal

manuscripts, and if so write synthetic type-2 variations for any groups which contain only one terminal manuscript as just defined. This routine must of course precede the routine for applying Greg's rules.

No other routines are required to disclose all the equivalents of

simple variations that lie hidden among the complex. It does not

follow, however, that the complex variations can thereafter be

ignored. If after the application of GREG'S rules it appears that two manuscripts are duplicates, examination of the complex variations may disclose that they sometimes disagree. Within the

province of the rules the two are effective duplicates, for the

replacement of one by the other will not alter the form of the

288




stemma, but the decision as to their exact genetic relationship will have to be made on some other basis.

It may also be that after the intermediaries have been determined

by GREG'S rules, an examination of the complex variations will show that some group or groups of manuscripts do not have their own common intermediaries. The larger group or groups in which the additional intermediary or intermediaries will have to be inserted will appear from GREG'S rules, but the exact place or

places of insertion will have to be decided on some other basis. It may be that when all is done neither simple variations nor

synthetic equivalents have been found. Then GREG'S rules will only apply within the largest mutually exclusive groups, whether

original or synthetic. The genetic relationships between these

groups will have to be decided on some other basis. These last three rules are like GREG'S in assigning a minimum of

intermediaries. However, with complex variations it is possible that two manuscripts might have had an ancestor between them and the hypothetical ancestor assigned them, which ancestors

might have sometimes differed in their readings. The problem is not a real one, since any internal evidence as to the existence of a

particular real ancestor would allow it to be located in the stemma and its readings to be deduced in the same way as those of a

hypothetical ancestor, while lack of such evidence would give no

grounds for speculation either way.

3.-Abnormal Agreements

If, with the limitations mentioned, complex variations can thus be made amenable to GREG'S rules, it is no longer necessary to assume the absence of abnormal agreements as a condition for the validity of the rules, for the effects of abnormal agreements can be negated by treating the variations that exhibit abnormal

agreements as if all the manuscripts which have not derived their

readings from their principal immediate exemplars have new

readings. The justification for so doing may be set forth more at

length in the following way: What happens when the various

processes of textual transmission operate? Multiform though these

processes are, they have only two results: either they reproduce in a copy the characteristics of its exemplar, or they introduce new characteristics. If these new characteristics are in fact to be found elsewhere, they are nevertheless new to the copy considered as a

Novum Testamentum, IX I9



VINTON A. DEARING

reproduction of its exemplar. A copy that has been corrected has characteristics that are new in the same sense. Likewise, a copy that has been made from more than one exemplar has new characteristics with respect to one of these exemplars. Therefore, when a critic identifies a reading as brought into some manuscript by correction, contamination, conflation, emendation, or chance, or

by some unidentifiable one of these causes, he may treat it as a new reading instead of letting it interfere with his tracing of descent. This treatment can be effected by rewriting the variations in more complex form.

Professor METZGER explains under the heading of "internal evidence" (pp. 209-2II, 217-246) how to determine which readings have been picked up abnormally in some manuscript or manuscripts. If from such evidence it is clear that, for example, the variation AB: CD has come about because of an abnormal agreement between A and B, the variation will be rewritten A: B: CD. It is not necessary to determine whether it was A or B which picked up the reading abnormally to see how to rewrite the variation.

Rewriting it instead of setting it aside preserves the evidence afforded by the remaining group or groups as to the lines of descent.

The mere fact that manuscripts divide over readings does not force the recognition of abnormal agreements in variations with

only one group of two or more manuscripts (e.g., A: B: CDEF: G) because the reading of the group can pass to all the manuscripts in the group by normal copying 1). This is not true of other variations, however. GREG showed that type-2 variations require the recognition of abnormal agreement when the groups in one variation

exchange manuscripts to become the groups in another, as in AB: CDE and AE: BCD. In this example, it does not matter whether B and E, or A and CD are thought of as exchanged, and

similarly in other cases. Complex variations may indicate abnormal

agreements similarly, and in other ways as well, as in AB: CD: EF and AF: BC: DE or in AB: CD: EF, A: BC: DE: F and AEF: B: CD. A general statement covering all the situations is that when two groups in one variation exchange manuscripts to become two

groups in a second variation, or to become a group and a synthetic group brought about in the second variation, or to become two

1) But MAAS, whom Professor METZGER cites first for the "classical method" (p. 157, n. i), accepts variations A: BC, AB: C, AC: B as indicating abnormal agreement.

290



NOTES ON GENEALOGICAL METHODS IN TEXTUAL CRITICISM 29I

synthetic groups brought about in the second variation, then abnormal agreement must be recognized.

Any manuscript not found in the groups in question in both variations is not involved in the abnormal agreement in any way. The rest may be thought of as forming a ring, linked together as

they agree together in these and any other variations. Most of the

agreements result from normal descent, one is an abnormal

agreement. It makes no difference if some or all of the manuscripts prove to be connected through intermediaries; the intermediaries when reconstructed will be found to have the same agreements. For example, variations AB: CD and AC: BD indicate that one of the agreements is abnormal. We can think of the manuscripts as

forming a ring

A-B I I C -D

in which each link is an agreement. Depending upon which is the abnormal agreement and on the other rules for constructing a stemma, the stemma may prove to be any one of the following or of a great number of other possibilities (the abnormal agreements are shown by dotted lines).

z

A Y Z Y Y I

I I I

I A i _ i- I i C--D A--B D C A--C D A B--D A B D--C

If the second example proves to be the correct stemma, the ring will prove to run through X instead of D and Y instead of C, and

similarly in the third (A X Y C), fourth (X B D C), and fifth

(Y X D C). The same ring is present in every stemma, then, and it can therefore be dealt with equally well in the simplest possible form.

In order to identify all the links in a ring and to treat one ring at a time, all variations in which the groups only pass manuscripts must be identified, but any new exchanges must be ignored for the time being. Thus ,if AB: CDE and AE: BCD are found, then if ABC: DE also occurs it must be analyzed with the first two unless



VINTON A. DEARING

ABD: CE also occurs, and so on, but AC: BDE, or ACD: BE will never be. With enough manuscripts and enough groups in the variations, one variation may have two or more sets of groups that

exchange manuscripts to become groups in other variations; these must also be dealt with one at a time.

If a group in one of the variations to be analyzed has more than two manuscripts, these manuscripts must be examined to see if

any has an agreement outside the group in another of these variations. For each such agreement, the rest of the manuscripts in the original group form a reduced group; after all the agreements have been identified, the reduced group or groups replace the

original group. Each remaining group and reduced group will constitute a link in the ring, and the ends of the links will be the

manuscripts this group has in common with different other groups. Thus, if the variations are AB: CDE and AE: BCD, each group is a link, as follows: A-B, B-CD, CD-E, E-A. If the variations are AB: CDE, ABC: DE and AE: BCD, the reduced groups BC and CD replace the original groups ABC and CDE because of the

agreement AE. The groups are then links as follows: A-B, B-C, C-D, D-E, E-A.

The correctness of the foregoing procedure is easier to determine

by experiment than by cogitation. The point is that an exchange of manuscripts indicates an abnormal agreement, but with more than four manuscripts it does not necessarily also make clear the links in the ring, which are primarily determined by the descent of the manuscripts. Below are two of the many possible stemmas with the rings just given, the first for the four-link ring.

x z

I I , r I B A-E DC B A--E D C

Both indicate that the abnormal agreement has proved to be AE, and that all texts stand alone in type-I or synthetic type-I variations. If the group AE does not mark descent, it does not come under GREG'S second rule, but the other groups do. In the first instance, the only other group of half or fewer of the manu-

scripts is AB, for which X has been assigned as the common inter-

mediary (the stemma shown would be chosen if X proved also to

292




be the archetype), but in the second instance there is also the group DE, which must have a common intermediary (the stemma shown would be chosen if Z proved to be the archetype). This additional common intermediary has introduced a fifth link in the ring. In the first stemma, the ring runs through X instead of B, and Y instead of CD, in the second it runs through X instead of B, Y instead of D and Z instead of C, but neither is essentially different from its simplest possible form, with which, therefore, we deal.

Further cogitation or experiment will show that if the manu-

scripts in any link share in all the abnormal agreements found in the variations, this link is the one which does not result from normal descent, if there is only one such, or cannot be distinguished from that link if there are more. The link is treated as though it did not exist by rewriting the variations so as to divide the manuscripts at each end of it and of any other link that cannot be distinguished from it, and to preserve the groups in the other links. Thus, if the variations are AB: CDE, ABC: DE, and AE: BCD, and the groups CDE and BCD have all the agreements identifiable as abnormal, then the manuscripts in the link C-D are the only ones to share in all the abnormal agreements and are to be divided. The first variation is to be rewritten AB: C: DE, not AB: CE: D, and the third is to be rewritten AE: BC: D, not AE: BD: C. To choose the alternate rewrittings would be to substitute new rings for old when the rewritten variations are compared in their turn with the others and themselves. Properly rewritten variations will be found amenable to GREG'S three rules after further processing under the three rules supplementary to GREG'S.

Any systematic procedure that leads to the identification of all the variations, rings, and links, and to the correct rewriting of the variations, will serve. This routine must precede the other two

already described. The variations that have been rewritten will

play no further part in determining the lines of descent for the

manuscripts, but since they alone record the abnormal agreements as such, they cannot wholly be forgetten without obscuring a part of the history of the readings.

B-Second Step: Determining the Archetype The rules for determining the archetype are simple and obvious.

If the readings identifiable as earlier can all be traced to a single point or line or contexture of lines in the linkage established by the



VINTON A. DEARING

foregoing rules, then the archetype lies here. The readings of an intermediary manuscript are determined by the normal agreements of the manuscripts between which it is intermediary. Thus, in Professor METZGER'S example the readings of X will be determined

by the normal agreement of any two of B, C, and D, or of any one of these and any one of the other four manuscripts; the latter type of agreement will determine the readings of Z as well, or of Y and Z if, e.g., B, and E agree. If no normal agreement occurs for a particular reading, e.g., if B, C, and D disagree with each other and all the others, the reading of the intermediary is undetermined: it might have been any one of the alternate readings, or different from them all.

Once the archtype is determined, the stemma is drawn by putting the archetype at the head and letting the linkage descend from it without breaking any of the connections. Thus, in Professor METZGER'S stemma for his seven manuscripts and the alternates to it, Professor METZGER'S stemma would be indicated if readings identifiable as earlier occurred in A and X but not Y, A and Y but not X, and X and Y but not A, and any other such readings in all three. The first alternate would be indicated if X was the only manuscript to exhibit all the readings identifiable as earlier; the second alternate would be indicated if this manuscript was Y. The third alternate would be indicated if the readings identifiable as earlier occurred in A or Z or both, but not always in both (that is, in A sometimes to the exclusion of X and Y, and in X and Y sometimes to the exclusion of A). The fourth alternate would be indicated if the readings identifiable as earlier occurred in B or X or both, but not always in both. If the readings identifiable as earlier were all exhibited by both B and X, then either the first or the fourth alternate would be possible, and the decision between them and still another alternate in which B would be the archetype would have to be made on other grounds.

If it is impossible to trace all the good readings to a single point or line or contexture of lines, then one of these is the principal descendant of the archetype and the others have picked up readings from the archetype by one or more of the processes that elsewhere result in abnormal agreements.

III.-SOME SIGNIFICANCES OF THE FOREGOING

A large part of constructing a stemma is thus routine, but not

294




all. Besides the fact that the textual critic has to decide which

agreements are abnormal, he may also have to decide a part of the essential linkage of the manuscripts on some other basis than the

principle underlying the rules given above. And he never resigns to routine the interesting task of evaluating readings on the basis of date and internal evidence, without which it is impossible to determine the archetype. The construction of a stemma will make

partly routine the reconstruction of the archetypal text, and will show the readings of which the reconstruction cannot be routine. But the construction of a stemma will not make routine the emendation of the archetypal text. On the other hand it is clear that with the rules and routines given above, an increase in the number of manuscripts to be dealt with makes the task of con-

structing a stemma harder only in proportion to the labor of

manipulating more symbols, it does not make it more obscure. If the form of the stemma is determined by the distribution of

the more original readings, the stemma cannot be cited as evidence for the originality of these readings without begging the question, so that Professor METZGER'S detailed explanation of how to evaluate

readings (pp. 207-219) needs some modification 1), but not to the

exclusion of genealogical evidence from this evaluation. The place for genealogical evidence is in the evaluation of readings for which there is no other evidence, as follows. Let us suppose that Professor METZGER'S stemma proves to be the correct one for his seven extant manuscripts, that is, that all the demonstrably more original readings in the seven can be traced back to Z and to no other point in the linkage. Presumably there will be some other readings where

originality cannot be demonstrated. It follows from the stemma that any normal agreement here between A and any other manu-

script, or between B, C, or D and E, F, or G is a reading of the

archetype, and therefore the more original. In other words, a

correctly drawn stemma makes clear the allowable generalizations

1) The same logic puts a particular burden of proof on the scholar who wishes to cite a "conflated reading" as being such (e.g., on Professor METZGER, who on p. 234 accepts xupLou Oeoo as a conflation of the alternate readings xuptou and 0eo5 apparently because it occurs "in Byzantine texts," and on pp. 236-237 accepts 0eo5 XpL9pro0 as the source of the alternate readings 0eou and Xptaro5 because it occurs in "the earliest and best Greek manuscripts" [and for other reasons]). The problem is analogous to that of whether Matthew and Luke each includes the whole of Q, or whether they include only independent but overlapping selections from Q.



VINTON A. DEARING

from demonstrably superior readings to readings whose superiority is not otherwise demonstrable. The great usefulness of such generalizations justifies the labor of constructing a stemma.

Professor METZGER groups the work of QUENTIN, GREG, HILL, and myself under the heading "statistical methods," a phrase adapted from Quentin (p. 163). QUENTIN and HILL do make

counts in the process of constructing stemmas, QUENTIN to

identify abnormal agreements, HILL to locate the archetype, but GREG and I do not. GREG in particular asserted that he had only clarified the logical implication and application of one of the

principles underlying genealogical methods and the fact that the

archetype must be determined by evaluating the readings on other

principles. The methods of STREETER and PASQUALI, which Professor METZGER groups under the heading "local texts and ancient editions" (p. I69) are also genealogical methods ). It follows that Professor METZGER'S second type of external evidence for evaluating readings, "the geographical distribution of the witnesses" (p. 209) is only a variant of the third, "the genealogical relationship of texts and families of witnesses," with which this discussion began.

Finally, Professor METZGER notes that a successful method for

quick determination of family relationships among manuscripts has yet to be devised (p. I8I). It may be suggested that once the stemma for the principal manuscripts has been worked out, others

might be placed by subjecting a random sample of the variant

readings to analysis by the rules set out above. Working out the stemma for the principal manuscripts however, is no brief task, and a representative sample of the readings in the other manu-

scripts would need to be fairly large, so I once again suggest that a

computer be used (in which case it would be easy enough to identify and process all the variant readings instead of just a sample). Professor METZGER doubts whether a computer "will replace the use of rational critical processes in evaluating 'good' and 'bad'

readings" (p. I69). P. TASMAN reported in I957 the successful reconstruction by computer of lacunae (up to five words) artifically introduced into texts from the Dead Sea scrolls 2), but the point

1) STREETER'S work on local texts preceded GREG'S Calculus and is therefore perhaps inevitably cruder in its analysis of genealogical evidence.

2) P. TASMAN, "Literary Data Processing," IBM Journal of Research and Development, I (I957), 249-256.

296




need not be pressed. It is enough to note here that comparison of

manuscripts is a thoroughly routine matter, and that the part of constructing a stemma explained in this paper is a simple logical process, which becomes routine once it is understood. There are in existence several computer programs for comparing texts and at least one (my own) for performing the routines outlined above for determining essential linkages. I am presently at work on a

program for determining archetypes from the evidence of the

"good" and "bad" readings, since this process is also routine once the readings have been evaluated by "the use of rational critical

processes." Professor METZGER'S explanation and demonstration of the "rational critical processes" of evaluating readings are

particularly full, clear, and instructive, allowing me in this paper to treat them as known to every reader or easily accessible. I have been concerned only to define more clearly the value of constructing a stemma and to explain more fully how to construct one by more

satisfactory methods than Biblical scholars have employed in the past.



Documents

Some Notes on Genealogical Methods in Textual Criticism