A DIFFERENT STEMMA FOR JOHN 7:53-8:11
Wilbur N. Pickering, ThM PhD
The criticisms of The Greek New Testament according to the Majority Text that I have seen have almost invariably fastened upon the genealogical reconstruction of John 7:53-8:11 (among other things). They note that sixteen times, out of 27 (where the MSS are significantly divided, as reported in the apparatus), the preferred reading is a minority one. Eleven of those times the minority is less than 30%! They discover that the discussion of the variant readings is redolent of Hort—minority readings are preferred against majority readings on the basis of internal evidence. Such a procedure sets aside the argument from statistical probability (which is usually associated with Majority Text Theory). They point out the discrepancy between theory and practice and wonder what went wrong. Can it be that when confronted with reasonably complete collations of the extant MSS the theory just won't work? In any event, why say the text is "according to the Majority" when it isn't?
In the Apocalypse we are presented with a situation where a clear majority is frequently unavailable. Where a majority reading does not exist we are obliged to use a minority reading, and defend our choice as best we may. John 7:53-8:11 is quite different—out of 33 significant variant sets only one lacks a clear majority reading. If the stemma offered on page xxv of the "Introduction" (H-F Majority Text) were incontrovertible then I suppose we should all loyally accept the consequences, but I find both the stemma and the discussion of the 21 variant sets to be less than convincing. Since the author of the "Introduction" recognizes that the stemmatic reconstruction (of Jn. 7:53-8:11) needs to be "searchingly evaluated", and this "calls for the cooperation of many minds" (p. xxxii), I venture to offer the following alternate reconstruction as my bit of "cooperation".
This article is based on a careful check of von Soden done by W.G. Pierpont (personal communication). My reconstruction of the text will be based exclusively on Soden's seven MS groups. Their relative size is as follows:
With this background I now present the 33 variant sets upon which this study is based. The total is sometimes 99% or 101% because I used whole numbers. Groups M4,3,2,1 are frequently internally divided, and their constituent MSS often go astray with a variety of added variants which are not recorded on the following chart, which is why the groups do not always add up to 100%. I use [ ] for 1-20%, ( ) for 21-95% and the whole # for 95+%.
I will start with some general comments about the MS groups. Based on the 33 significant variant sets (significant for the reconstruction) we observe the following:
- 7 is never alone, except for the one time when 5, 6 and 7 go separate ways (#24).
- 7 and 5 are the only groups that never divide (according to von Soden).
- 5 and 6 each stand alone against the rest of the stemma—a condemning circumstance. 6 does so three times (3, 21, 22), plus two more times where it influences group 4 (2, 4). 5 does so four times (5, 7, 8, 32), plus four more times where it influences group 1 (11, 15, 19, 26).
- 5 and 6 (entire) never agree (where one of the three major groups diverges)—they are the extremes (which makes it unlikely that either one stands closest to the Autograph).
- 3,4 and 1,2 evidently form groups at the next level up, which are themselves closely related.
- 1 is fairly close to 5 but far from 6, while 4 is just the opposite.
Going into more detail, I will start with groups 1-4. It is obvious that they are by no means monolithic, and one has to wonder if Soden's assigning of individual MSS to these groups was altogether felicitous. Ignoring solitary MS deviations from a group:
4 divides 14 times (2 being three-way splits),
3 divides 10 times (2 three-way),
2 divides 21 times (5 three-way and 3 four-way!),
1 divides 24 times (7 three-way),
within our 33 variant sets—they each have further divisions.
However, it seems clear that four such groups do exist and it is generally possible to determine the reading of the exemplar. So I next ask how these four groups interrelate:
It is evident that 2,1 form a group and 4,3 form a group; that together they form a larger group is clearly demonstrated by the next chart. The seeming closeness of 3,2 is because they are the steadier members of their sub-groups, 4 and 1 being more erratic (4 has five singular readings plus assimilating to 6 three times; 1 has two singulars plus assimilating to 5 five times; while 3 and 2 have none). Where 3,2 agree they preserve the reading of the grandparent. The parent of 2,1 I call j, and that of 4,3 I call k. The parent of j and k I call h.
The following chart makes it easier to see the pattern. The reading of 7 is always "x"; if 6 differs it is always "y"; if 5 differs it is always "z". "w" is used for any variant distinct from the first three. Brackets are used when a variant is clearly derived from another. Braces are used for a second derived variant. Parentheses are used to indicate separate groupings. Backward slant lines are used when a grouping is anomalous. A question mark means that I don't know what is going on. The "=" column gives the reading followed by h, and demonstrates that h had a mixed text.
The crucial question for our stemma is the interrelationship of the four main groups: 7,6,5 and h. That h follows "x" 13 times, follows "y" 6 times, "z" 4 times and "w" 3 times shows that it has a life of its own and must be treated as a separate group. We observe the following:
- 7 and 5 agree 17 times (ignoring #18).
- 7 and 6 agree 14 times (ignoring #18), plus six further times where a significant part of 6 goes with 7 and 5 (53%, 46%, 45%, 39%, 38%, 35%).
- 7 and h agree 14 times.
- 6 and h agree 11 times.
- 5 and h agree 9 times.
- 5 and 6 (entire) never agree (except in #18 where 7,6,5 all agree—I included #18 as an added illustration that h has a life of its own).
So, 5 and 6 are the extremes and 7 and h are candidates for mixed texts—and yet they do not depend directly upon each other, for 7 and h never agree unless 5 or 6 is with them. However, 6 and h agree alone two times (+ eight further times when one or the other is divided), while 5 and h agree alone four times; further, h splits, going with both 5 and 6 four times. Since 7,6,h agree nine times and 7,5,h agree five times (+ four further times when h is divided), we must posit nodes above 5 and 6 but which are separate from 7. A careful scrutiny of h makes clear that it is a mixed text, drawing from the exemplars of 5 and 6. Since 6 has seven singular readings and 5 has five it is clear that each has corrupted its exemplar and neither can stand closest to the Autograph. 7 gives no evidence of being mixed; it has only one singular reading, precisely in set 24 where each of the four main groups has a different reading.
I conclude that 7 is independent and at the same rank as the exemplars of 5 and 6. This accords with a normal transmission where rarely will more than one copy at a given rank create a variant at the same point, and almost never will two make the same mistake independently (except for very easy and common transcriptional ones).
Perhaps a design will help to visualize the options. Consider:
If 7 is viewed as a mixture of 5 and 6 the even selection is strange, plus a total lack of conflations. The six splits in 6 plus the fact that 6 has seven singular readings and 5 has five singular readings, while 7 has only one singular (precisely where all three disagree), point to a) as the best interpretation.
Since h has obvious secondary readings, including two conflations, plus a deal of mixture, b) seems to be the required interpretation. Evidently h drew more heavily from the ancestor of 6 than from the ancestor of 5.
I now offer a stemma that I believe matches and accounts for the statistical evidence just described. The Autograph is represented by "A". The letters f,g,h,j,k designate reconstructed archetypes. The numbers within nodes are those of Soden's MS groups. The numbers beside the nodes refer to the variant sets in which the node introduces an error (such errors vindicate the positing of a node, especially the reconstructed ones): thus f introduces five errors, 7 introduces none, g introduces thirteen, etc. The solid lines indicate linear descent; the broken lines indicate occasional influence (smaller dashes = lesser influence). The symmetry is arbitrary, as is the chronology. I am not concerned to defend the chronology, but I do happen to think that it is "in the ballpark"—a possible, even reasonable, approximation.
Before returning to the variant sets to discuss transcriptional probabilities, I wish to comment on the stemma. 7, f and g are independent of each other, yet only one abandons the original at any point (except for #24 where two of them [or all three] do, in separate directions). This is just what we would expect in a normal transmission, where the archetypes at a given rank are independent (of each other). With the exception of #24 there is always a clear majority, of which 7 is always a part. I consider that 7 has faithfully preserved the Autograph, including in #24 (see the discussion of transcriptional probabilities below). f is tolerable, but g is a maverick, as is 5. h has a mixed text. (The placing of "Latin" and "Egypt" is deliberate: it appears to me that 6 and 4 reflect "Egyptian" influence, whereas 5 and 1 reflect "Latin" influence.)
And now for the probabilities.
The variant sets are taken in linear sequence, beginning in 7:53:
1) 7 and g preserve the original; 6 influences 4. f creates a variant and influences h; j changes h. Presumably the plural forms were unthinking assimilations to the Subject in verse 52. The reading of 3,5 is possibly a harmonization with the same verb in the next line.
2) 7, f and g preserve the original. 6 creates a variant and influences 4. Since the main participant or focus changes from 7:53 to 8:1, the de is required (a norm of Koine discourse structure). 6 has an inept stylistic change. 4 follows 6 but drops the article.
3) 7, f and g preserve the original. 6 creates a variant. We may never know what got into 6 here and in the next example. In Luke 24:1 it really was "very early" (still dark) but here there were already people in the temple. That "very" separates the adverb "again" from its verb is awkward. Anyway, it is scarcely credible that 6 could be right against the whole stemma.
4) 7, f and g preserve the original. 6 creates a variant and influences 4. Since there is no ambiguity the repetition of "Jesus" is unnecessary, if not grammatically bad. 4 seems to have replaced the verb of his exemplar with that of 6.
5) 7, f and g preserve the original. 5 creates a variant, as does (4). Just why 5 dropped the prepositional phrase here is hard to say, but that is not sufficient reason to prefer it against the whole stemma.
6) 7 and f preserve the original; g creates a variant and is followed by 6 and h. 7 influences (6); 5 influences (1); if (3) was influenced by 7 or 5 it is the only such case—perhaps it managed a lucky gloss on its own. The omission was probably thought to improve the style. For someone to supply the words would also be easy. External evidence must decide the issue.
7) 7, f and g preserve the original. 5 creates a variant, attempting a stylistic, or grammatical, "improvement".
8) 7, f and g preserve the original. 5 creates a variant and influences (1). This variant evidently goes with the preceding one. I see nothing sufficient to overturn the external evidence.
9) 7 and f preserve the original; g creates a variant and is followed by 6 and h; 5 influences j. The article may have been thought to be stylistically better. When mesoV is part of a prepositional phrase and is modified by a genitive construction no article occurs; without the modifier an article frequently occurs, but not always—as in v. 9 below. External evidence must decide.
10) 7 and f preserve the original; g creates a variant and is followed by 6 and h; 5 influences j. This verb is parallel to the one that begins v. 3, so the present tense is appropriate. The aorist tense is presumably a superficial assimilation to the participle a few words before.
11) 7, f and g preserve the original. 5 creates a variant and influences (1). The gloss seems officious. I see nothing sufficient to overturn the external evidence.
12) 7 and g preserve the original. f creates a variant and influences h. This set and the next two go together. The change here was deliberate, but it is hard to know what the motivation may have been. Perhaps the demonstrative pronoun was felt to be too scornful.
13) 7 and g preserve the original. f creates a variant and influences 5; h corrupts f; 1 conflates j and 5. The confusion within h does not inspire confidence. 5 could represent a stylistic "improvement", making the verb (of f) more graphic (in which case 5 would create the variant and h reproduce f).
14) 7 and g preserve the original. f creates a variant and influences h. The choice here is controlled by the choice in 13).
15) 7, f and g preserve the original; k corrupts h. 5 creates a variant and influences (1). This could have gone either way, though perhaps "our law" is more typical of John. External evidence decides.
16) 7 and f preserve the original; g creates a variant and is followed by 6 and h; 5 influences (1). The passive seems more appropriate. I would say that our Lord's subsequent use of the words liqon baletw reflects precisely the scribes' liqoboleisqai.
17) 7 and f preserve the original; g creates a variant, followed by 6 and h; 7 influences (6); 5 influences (1). The absence of the phrase is stylistically and semantically better. An officious but not very perceptive copyist thought he was filling in implied information, but "her" does not match the plural "such".
18) 7, f and g preserve the original. h creates a variant, as do 4 and 1. h's exemplar may have had a weak cross stroke in the epsilon so he read it as a sigma.
19) 7, f and g preserve the original. 5 creates a variant and influences 1. The infinitive is the more awkward form, sufficient reason for modern critics to prefer it, but the external evidence seems clear.
20) 7 and f preserve the original; g creates a variant, followed by 6 and h; 7 influences (6); 5 influences (1). Is it a gloss or an omission? It could have happened either way. Perhaps g felt that "not paying attention" made Jesus seem impolite, especially considering the prominence of His interrogators.
21) 7, f and g preserve the original. 6 creates a variant; if it influences 2 it is the only time—(2) may have made the same stylistic change independently, having done something similar in the preceding sentence. The heightened form was probably a stylistic "improvement". In any event, it is scarcely credible that 6 could be right against the whole stemma. Notice that (2) also heightened the verb in the preceding clause.
22) 7, f and g preserve the original; h corrupts g; 5 influences 1. 6 creates a variant. It is clear from the text that Jesus squatted and stood up twice, but 6 has Him squatting twice without standing up in between—if a scribe didn't notice that Jesus squatted twice he might have decided to remove the "inconsistency" of His standing up twice. The finite form could be a stylistic retouching.
23) 7 and f preserve the original; g creates a variant, followed by 6 and h; 5 influences (1). A toss-up, although the prepositional phrase seems to be more dramatic, delightfully appropriate.
24) This is a tough one, the only time that the three major groups go separate ways. That is doubtless why this set has the most variants and confusion. There are three variables: the Case of the pronoun, the article, and constituent (word) order. They all have different alignments:
authn—(1),(2),4,5,6 VS auth—(1),(2),3,7
liqon—(1),(2),4,6 VS ton liqon—(1),(2),3,5,7
IO,V,DO—(1),2,(3),4 VS IO,DO,V—(1),5 VS DO,V,IO—6 VS DO,IO,V—7
The line-up is the same for the first two variables, except for 5 which switches. It goes with 6 to give the accusative majority status, but goes with 7 to give a majority to the article. The constituent order seems to be hopelessly confused, but it may be noted that there is a clear majority for the direct object in the first position (and therefore for the relative sequence DO,IO within the clause) and for the verb in the last position (which leaves the middle position for the indirect object). But how can we explain the fluctuating alignments?
I reconstruct the sequence of transmissional history thus:
Being freely interpreted this means:
- 7 preserves the original.
- g, followed by 6, creates a variant: changing the case, dropping the article and inverting IO and V.
- f creates a variant, inverting DO and IO.
- 5 modifies f, changing the case (independently —epi takes both).
- h, followed by j, k and 3, creates a variant (building on f), moving DO to last place (while preserving case and article).
- 4 changes the case and drops the article, presumably influenced by 6 (a frequent occurrence).
- Four of the sixteen M3 MSS invert V and DO (while retaining the case and the article); another three assimilate to 6.
- 2 splinters: 28% staying with j, 16% changing only the case, 42% changing the case and dropping the article, with 14% going beyond the 42% to invert V and DO besides. (Again one wonders if von Soden correctly assigned some of the MSS.)
- As for 1: 48% retain the case and article of j but invert V and DO, presumably influenced by 5 (a frequent occurrence); 43% retain the word order of j but change the case and drop the article (presumably influenced by 6, since two M1 MSS assimilate to 6 completely).
This reconstruction gives a reasonable explanation for the peculiar range of attestation for the Dative case. It also accounts for the range of attestation for the article. In fact, it pretty well accounts for the whole snarl of variation. M7 preserves the original at every point.
Assuming that 7 is right all the other times, it has far and away the best "credibility quotient". If we start with 7, 6 would be a stylistic "improvement", dropping the article and placing the indirect object last (perhaps he took "first" to be modifying "stone", as a number of prominent English versions have done). Why f would put the indirect object first is not clear, unless it be for emphasis. h devised a new constituent order, which was retained by all its descendants, except for (1) which was influenced by 5. Considering the fluctuating alignments throughout the variant sets it would appear that copyists frequently had access to more than one exemplar and were not above picking and choosing—it seems reasonable to suppose that they would be most willing to do so when confronted with confusion in the tradition.
The split of (1),(2),3 for the article and (1),(2),4 against it is stemmatically anomalous—I feel obliged to conclude that a lot of picking and choosing took place in this case. The same split for the dative versus the accusative is also anomalous, although since epi takes both cases this particular change could have happened independently more than once. It is not my custom to prefer "harder" or more awkward readings, but perhaps this is an appropriate place to cash that check. If 7 is stylistically more awkward than some of the other variants then it may have given rise to them.
25) 7 and f preserve the original; g creates a variant, followed by 6 and h; 7 influences (6); 4 creates a variant. Surely it would be easier to omit a clause of this sort and size than to invent it, especially if it occupied just one line. 4 is wild; speculation is pointless.
26) 7, f and g preserve the original. 5 creates a variant and influences 1. The omission could be the result of homoioteleuton. In any event, 5 could scarcely be right against the stemma.
27) 7 and f preserve the original; g creates a variant, followed by 6 and h; k corrupts h and 5 influences j. A toss-up, unless it be that g dropped "only", followed by h, and in subsequent efforts to fix it (6), (4) and (2) replaced it but in the wrong order. In that event 7 influenced (6).
28) 7 and f preserve the original. g creates a variant, followed by 6 and h; j omits. The solution here involves the following set as well. Faced with a lacuna before eipen gunai, if someone felt the need to create a gloss like the longer reading would he also change "woman" to "to her"? It is much more likely that he would supply something like the shorter variant and leave "woman" intact. But on the face of it the longer reading is not the sort that a copyist would concoct out of nowhere. If we start with the longer reading it is easy to see where someone could be troubled by the seeming contradiction—it says that Jesus "saw nobody" except the woman when in fact there was a crowd taking in the proceedings. So he "solves" the problem by replacing it with the innocuous shorter reading; but now he has two pronouns too close together, so he changes the second one to "woman". It may be instructive to combine sets 28) and 29):
The descendants of g are evidently confused, which tends to diminish their credibility. Whether the longer reading was accidentally dropped (if it occupied a full line) or deliberately changed, there is no "internal evidence" sufficient to set aside the clear external evidence.
29) 7 and f preserve the original; g creates a variant; h conflates f and g; 5 influences 1. (3) drops the pronoun independently. See the discussion above.
30) 7 and f preserve the original; g creates a variant, followed by 6 and h, which (4), (2) and (1) omit. 7 influences (6). If the shorter reading were original, who would have thought of adding the demonstrative? The shorter form is very nice the way it is! If the longer reading were original someone might well have felt that the demonstrative was scornful and unbecoming. Or the demonstrative could have been dropped through parablepsis. (4),(2),(1) evidently perpetrated a further omission, whether wittingly or not is impossible to say.
31) 7, f and g preserve the original; k corrupts h (varying the word order but keeping the pronoun). 5 creates a variant, and influences j. The external evidence for the word order is overwhelming. 5 dropped the pronoun, either by accident or as being superfluous.
32) 7, f and g preserve the original. 5 creates a variant. This omission by 5 has little to recommend it.
33) 7, f and g preserve the original. 5 creates a variant and influences (1); 4 makes the same mistake. The prepositional phrase could have been thought to be redundant, or lost through parablepsis. Presumably 5 and the archetype of 4 omitted independently; later (4) also dropped "and".
I confess that I find discussions of probabilities to be wearisome and frustrating. It is seldom possible to rise above mere speculation. However, perhaps we can glean a few reasonably solid tidbits from the foregoing exercise. I would say that in #2 group 6 is clearly wrong, grammatically bad. In #3 group 6 is unacceptable; it is also inferior in #4. In #11 and #12 the reading of 5 is unacceptable. In #16 group 6 is unacceptable, as also in #22. In #32 group 5 seems inferior. Group 4 is erratic, but is not a serious contender in any case. Group 1 is badly splintered, and is not a contender. I submit that our exercise confirms that neither 6 nor 5 can stand closest to the Autograph.
So, what have we learned with reference to our larger concerns? To my mind, rather than constituting an embarrassment John 7:53-8:11 confirms the validity and workability of Majority Text Theory. The transmission was "normal" in the main. In a normal transmission the copyists of a given generation would almost never commit the same error independently, and seldom would more than one introduce an error at any given point—thus the true reading should have majority attestation at all levels. In spite of a rather high incidence of variation (considering that "Alexandrian" and "Western" witnesses were excluded to begin with—thereby avoiding most of the results of the "abnormal" transmission) the pericope vindicates our expectation. Even though it evidently passed through more turbulence than any other passage in John's Gospel, still the original text is attested by a clear majority of the MSS.
To conclude, if the essential thrust of my discussion is convincing to my peers, I earnestly recommend that in any future reprinting of the H-F Majority Text the text of John 7:53-8:11 be corrected to reflect the majority attestation. (The same holds for the R-P Byzantine/Majority Textform.) It would also be necessary to revise the corresponding discussion in the "Introduction". To the extent that our theory is viewed as a threat to the Establishment it will certainly be attacked, that we know, but I see no virtue in giving them gratuitous ammunition.
Robinson-Pierpont continue to misunderstand (p. 494) what Hodges-Farstad and I mean by "stemma". We are not talking about the genealogy or descent of MSS; we are talking about the genealogy or descent of readings. Soden's MS groups are based on "profiles" of readings in common, a concept that R-P seem to accept; a concept that seems to me to be obviously valid, and necessary. I think we all would agree that "genealogy" as applied to MSS is unworkable.
But see chapter 7, "Examples and Implications" and "Conclusion" where I argue that demonstrated MS groupings and relationships supersede the mere counting of MSS.
Robinson-Pierpont claim that both 5 and 6 read "the sins of each one of them" at the end of 8:8, because Soden seems to indicate this in two places (II:427 and I:514). However, in two or three other places where Soden should mention it, he does not. Both UBS3 and N-A26 would appear to disagree with R-P (although the text of both editions is virtually identical, being the work of the same five editors, I understand that the apparati were prepared separately). UBS3 uses Byzpt to refer to percentages of MSS ranging from 74% to 24% (in the Pericope), M5 and M6 each receiving this classification, when alone or with scant company. In 8:8 the gloss does not rate even a Byzpt, the absence of the gloss rating Byz. Since M5 + M6 would represent 58% of the MSS the gloss would have to rate Byzpt—it follows that UBS3 does not agree with R-P. Although the use of pm and al in N-A26 is not consistent, that the gloss rates only an al would seem to indicate that N-A26 also disagrees with R-P. (Both codices U and P have the gloss, which is a curious circumstance in that U usually goes with M6 and P with M5. Perhaps one of Soden's assistants took the two codices as representative of the two groups without really checking out the bulk of the MSS.) [Maurice Robinson has recently (1998) completed a collation of 1,635 MSS for the Pericope and Pierpont (personal communication) informs me that a small minority of MSS, from groups 5 and 6, have the gloss. Robinson's collation confirms that 5 and 6 (entire) never agree (where one of the three major groups diverges).]
Both H-F and R-P seem to accept Soden's equating of M7 with his Kr group, which is defined as the later Byzantine text, as distinct from the main tradition (Kx). I am inclined to suspect that Soden's judgment derives from his presuppositions far more than from the evidence. I believe that an independent review of the evidence will show that M7 is really ancient. An analysis of the collations in Acts published in Text und Textwert and in Luke 1, 10 and 20 done by F. Wisse shows that f18 (alias Kr) is both independent and ancient. f18 enters into a considerable variety of shifting alignments, quite distinct from Kx, and frequently shares readings with ancient MSS, Versions, Fathers (against Kx), but the sharing and aligning are not systematic, are not predictable—therefore f18 must be both independent and ancient. In fact, the collations done by Robinson (see footnote 3) have led him to affirm that all the major groups are ancient and independent—that is, he did the "independent review of the evidence" and arrived at the conclusion I predicted: M7 is ancient (3rd if not 2nd century).
The New Testament in the Original Greek according to the Byzantine/Majority Textform contains an appendix about the Pericope Adultera, pp. 494-505. R-P print various forms of the pericope including that of M7 (my choice), the H-F Majority Text and their own Byzantine/Majority Textform. They state that they generally prefer M5, consider M7 to be inferior to both M5 and M6, and yet M5 + M7 always outweigh M6 (pp. 496-97). When we inspect the R-P reconstruction, however, we discover the following: they reject the reading of M5 nine times and place it in doubt five further times; they place M5,7 in doubt twice (against their stated principle); they follow M5 (alone) without question only once—in fact, their reconstruction is closer to M7 than it is to M5. It all leaves me a little puzzled: why then do they say that M5 is the best and M7 the worst among the three major groups?