Friday, February 06, 2009

The More Difficult Reading?

In a comment to the post on Phil 1.11 Andrew Wilson raised an interesting question/issue. With his permission I have copied it here:

To put the problem of evidence for lectio difficilior into context, Royse's and Head's studies showed the ratio of singular omissions to singular additions among early papyri running at about 75:25 (in percentage terms, or, for every three omissions you get one addition). That sort of evidence, to me, is incontrovertible - scribes omitted far more often than they added.

But, my figures for lectio difficilior (based on over 2000 singular readings) show a ratio of over 200 harder readings to about 10 easier readings (20:1), or putting it into percentages: 95% harder readings to 5% easier. Those sorts of ratios, to me, put lectio difficilior out of business.

However, when I mention these sorts of figures to textual critics, most instinctively (think knee-jerk reaction) spring to the defence of the canon. The problem is that some of these defences are poorly-thought-out, almost as if the canon must be defended at all costs.

Let me give two examples:

1. Easier readings would be more attractive to subsequent copyists, and more likely to be perpetuated, whereas harder readings are more likely to be later corrected. Therefore, the fact that harder readings predominate among singulars means little because the long-term effect on textual transmission would be in the opposite direction. To me, this objection (a) is circular reasoning, for it assumes what it should be proving (that scribes try to improve the text, rather than introduce further difficulties), (b) is not based on any evidence - the question of at what point a reading becomes so difficult that it tempted scribes to alter it should be one for evidence to determine, not armchair speculators, and (c) it puts the canon beyond the reach of any contrary evidence. It is like the story of the man whistling in Chicago who said he was whistling to keep the tigers away. When it was pointed out that there were no tigers in Chicago, he replied: 'See, it works'.

2. A second defence involves introducing the issue of intentionality by claiming that lectio difficilior only applies to intentional changes. This defence, likewise, brings with it questions and problems. (1) It apears to commit the fallacy of the excluded middle - it ignores the possibility that a reading might have been created in more than one stage, as a result of a wandering mind or as a salvage reading of an earlier nonsense reading. (2) Even if we cull nonsense readings and cases of hom. from the count (as likely to be unintentional changes), we are still left with easier readings as an extreme minority, for the great majority of singular readings remaining are neutral or harder. This leaves the 'nuanced' defender of lectio difficilior with a conundrum. Either intentional changes to the text rarely seem to result in an improved text or, alternatively, the vast majority of textual changes must have been unintentional (even though they appear otherwise). Either way, the objection (and lectio difficilior) fails. In short, the objection does not dispose of anywhere near enough evidence to stand.

And this is where I would like some help ... Are there any other good defences of lectio difficilior that I should be considering?


  1. just wondering, with no proof whatsoever, that if we abandon lectio difficilior altogether, we end up with Byz type of text ... do we?

  2. Well, one defence might be to point out that scribal habits deduced from singular readings may not be all there is about those habits.

  3. "In a comment to the post on Phil 1.11 Andrew Wilson raised an interesting question/issue."


    "if we abandon lectio difficilior altogether, we end up with Byz type of text ... do we?"

    No, we'd have to eliminate all the harder readings from the Byz text too, and come up with a text that doesn't exist in any single mss.

  4. Daniel, I did not say Byz text, but Byz type of text :) There is a distinction. I was thinking conflation and harmonisations etc...

    I think we would indeed end up with a text that does not exist in any manuscript, but even the Byz text is like that (to some degree).

  5. I don't think that abandoning lectio difficilior would result in a very different text to what we have at present. Abandoning lectio brevior (particularly in the case of small add/omits) would have a far more significant (cumulative) effect upon the text, and it would probably be slightly more Byzantine.

    On the other hand, abandoning lectio difficilior would probably do little more than reduce the amount of journal articles arguing for eccentric readings based on very few mss. Ok, there probably would be other consequences ...

  6. Okay,just to put things in focus, I count 174 singular readings in Matthew 20, based on Swanson, Von Soden and Tischendorf.

    Of these, over 20% are nonsense readings, one-third are neutral changes (e.g., substitutions of kai for de, transpositions, etc), and 37% are stylistic changes (e.g. additions/omissions of pronouns, conjunctions; and substitutions that do not affect the sense, but say it in a slightly different way).

    This leaves us with just under 10% of singular readings, which are all harder readings (there are no easier readings in this chapter). Here are all 15 of the harder readings:

    (20:8) X(Tisch) omits ewj twn prwtwn
    (20:19) X(Tisch) omits kai staurwsai (Christ's resurrection presupposes his death)
    (20:26) 71(vSod) omits megaj
    (20:3) D: diexelqwn instead of exelqwn
    (20:12) 1346: umin for hmin
    (20:14) 01: su for soi
    (20:15) 1071: exesth for exestin : ‘he was amazed’
    (20:15) 69: mou for sou
    (20:16) D(037): klhtai for klhtoi (females are called, but only males are chosen?)
    (20:19) 124: autouj for auton
    (20:21) E*: autoij for auth
    (20:26) 579: oude for outwj
    (20:26) 579: hmin for umin
    (20:32) 1588(vSod) - anastaj for staj (Jesus did not need to 'stand up')
    (20:34) 01*: autou for autwn

    Now, why is it that this is the way it looks wherever we wish to go in the NT: a good number of cases of harder readings, but very few cases of scribes actually improving the text?

  7. Ok, a guestion, Andrew. Why are you thinking about singular readings? Though Royse's study shows something foundational about the scribal habits on the basis of singular readings, I do not think that these perceived habits are watertight guidelines in making text-critical decisions. IMO.

    Secondly, singular readings may say little about harder readings that are not singular or sub-singular. IMO.

    I agree that abandoning lectio brevior would result more changes, but abandoning lectio difficilior would also alter many things.

    lectio difficilior is not looking for the hardest reading, and it cannot be applied mechanically, but it is an important argument, which should not be abandoned altogether. IMO.

  8. A distinction should be made between singular "difficult" readings and those "difficult" readings otherwise shared among a number of witnesses which lack indication of accidental or individual editorial corruption.

    And yes, the Byzantine Textform itself retains its own share of readings that are more "difficult" than those found within the current critical text; thus, I have expressly retained the canon of the "more difficult reading" as part of my own theory and praxis.

  9. Don't most (see Griesbach, for example) restrict the "harder reading" rule with many qualifications? For one, almost none of the editors before Lachmann (Origen might be an exception, but still he preferred "majority" readings), unless I'm mistaken, none would have chosen a singular reading no matter how difficult it happened to be, this due to prudent qualifications to a very logical rule.


  10. Timo raises one objection, at a more fundamental level, to studies using singular readings. Timo seems to argue (forgive me if I am misreading you Timo) that singular readings are not a sufficient way to determine scribal habits.

    Barbara Aland’s criticism of Royse’s work took this line. Barbara Aland, in a number of studies, looks at the scribal habits found in certain early papyri by examining deviations from the ‘standard text’ (i.e. NA). She is perhaps right to argue that singular readings do not tell us everything (or enough) about an individual scribe or manuscript.

    However, there are two problems with Aland’s argument. Firstly, it is based on the question-begging assumption that certain texts or mss are roughly equivalent with the original text. Aland’s assumption is even more problematic when we realise that the ‘standard text’ is based on transcriptional canons like lectio brevior and lectio difficilior in the first place. How can we test transcriptional canons by comparing mss against a text that has an inbuilt bias for exactly the type of readings we are trying to test?

    Singular readings offer a more objective basis for identifying scribal errors, as they are the readings most highly-likely to be scribal errors.

    Secondly, Aland misunderstands Royse’s work, for Royse was not interested in individual scribes per se (as Aland was), but with scribal habits in general. Whilst the few singular readings that we might find in an early papyrus (a la Aland) are not enough to tell us much about one scribe’s habits, once we start collecting hundreds of singular readings from a spectrum of mss, we have plenty of data to start working on. And this information gives us the only solid, objective evidence we have about scribal habits.

  11. Dr. Robinson’s objection was slightly different. He argued, if I might paraphrase him, that singular readings are a different genus of variant reading from any other.

    There is, again, some limited truth to this objection, insofar as nonsense readings (and cases of hom.) do not make it into ‘normal critical apparatuses’ which show us readings that should possibly make it into the text.

    However, once we eliminate these sorts of readings, we are still left with about 75% of singular readings. And these singular readings are EXACTLY the same as any other variant readings that are to be found in any critical apparatus.

    Thus, 10% of singular readings are transpositions, a very large percentage of singular readings are omissions and additions of single words, there are large numbers of perfectly sensible, neutral, subsitutions among singular readings. And there are harder readings.

    The only difference between singular readings and readings with other support is the 'other support'.

    To prove this, have a look at Von Soden’s apparatus, which is ordinarily split into three sections, the highest section comprising the readings most worthy of notice and the lowest section comprising (supposedly) the most irrelevant bits of textual flotsam.

    I have opened Von Soden at random at his page containing John 3:32 – 4:7. Here are the first few readings in the top apparatus (he only has two apparatus sections on this page): add kai, omit touto, ekeino for touto, add de, omit gar.

    Why do these readings get put in the upper apparatus? Why are they not relegated to an apparatus for ‘textual flotsam’? For one reason only: they are found in a substantial number of mss or in witness that Von Soden thinks (based on external evidence) to be significant, whereas the lower apparatus contains readings found in fewer mss.

    In other words, Von Soden’s apparatus distinguishes between readings – not on some intrinsic basis – but solely in terms of the external support.

    Thus, I would argue that the claim that singular readings are ‘different in quality’ is based on thin air. On the contrary, singular readings (when studied in sufficient numbers) give us the best and most objective guide to what scribes actually did when copying the text, and thus they are our best guide to what transcriptional canons should look like.

    If we don't base our transcriptional canons on scribal errors, what should we be basing them on?

  12. "almost none of the editors before Lachmann would have chosen a singular reading no matter how difficult it happened to be, this due to prudent qualifications to a very logical rule."

    There is no rule against choosing a singular reading--NA27 does it, anyway. I suspect, thought, that it only does so WHEN it is a difficult reading.

  13. To make it clear: I agree with both Wilson and Royse that singular readings can be used to establish a pattern of scribal habits; this in fact was the method employed in my 1982 doctoral dissertation ("Scribal Habits among MSS of the Apocalypse").

    My earlier comment was merely to suggest that "difficult" readings that are shared among several witnesses represent a different category for analysis and evaluation.

    However, in cases where such shared readings also correspond with the proclivities of a particular scribe -- in such cases the testimony of that particular MS becomes suspect, and should not be used as strong evidence in conjunction with other MSS that retain the same "difficult" reading, but whose scribes do not reflect the same proclivity.

    Interestingly, part of B. Aland's criticism of Royse includes her view that singular readings simply don't exist, and therefore such should not be used as a basis for evaluating scribal habits. How does she make such a claim? On the (dubious) ground that numerous MSS no longer extant may have read such otherwise "singular" readings -- a very convenient and generally fallacious argument ex silentio in my opinion.

  14. I admit it sounds circular. The hardest reading is to be preferred especially when the best and most varied MSS contain the harder reading. How do we know what MSS are the best? We know which ones are best by their high amount of more original, harder readings. But actually, now, we know what MSS are best by how well they align to NA27, the new textus receptus.

    So Andrew, in Luke 24:37, we have 3 readings listed in NA27: (1) QROHQENTES (p75 B); (2) FOBHQENTES (Aleph W); (3) PTOHQENTES (the rest).

    Can the harder reading canon play a role? Two (2) is the easier reading and has only 2 MSS in support, so is rejected out of hand. One (1) is relatively rare (used only 3 other times), and so automatically hard, but three (3) is used even fewer times (only once) and only in Luke. It is the harder reading while at the same time uniquely Lukan.


  15. Steven Avery9/05/2009 1:56 am

    Hi Folks,

    Since it is clearly used in a circular fashion, wouldn't it be better to go back to when lectio diffilior was enthroned as a major Bible textual concern and why ?

    This became a handy argument point in 1880 for the alexandrian manuscripts which were filled with ... harder readings .. (which in some cases is a fancy word for blunders). Without lectio could Scrivener et al be convinced to take virtually every Aleph-B reading ? Likely not.

    Now it is clear that to the Reformation Bible giants lectio difficilior was a strange concept that should never be used to support the unevangelical corrupt reading. (The Jan Krans discussions are interesting but do not change the basic fact.) In fact, at base lectio difficilior abused (which it always is today) is simply incompatible with evangelical belief. Although it does make for the cute situation of competing opposite theories from the same data. (pointed out by Professor Robinson and others)

    Maybe you could find a few cases where the Reformation scholars seemed to directly consider the concept. However a geography blunder (e.g. Gerash) would be seen as the result of an ignorant scribe far away from the turf -- rather than a savvy scribe correcting Lukes original faux pas.

    As to NA27 being the new TR, naaahh.. to many it is a textus dictatorius, or a textus inconsequentialus -- and the Christian believer who sees how the sausage was made spit it out long ago.

    Steven Avery