Loading ...
Sorry, an error occurred while loading the content.

A challenge

Expand Messages
  • Vinton A. Dearing
    Dear listers: A textual stemma has a general form that may need to be determined before the stemma is decided upon. For example, with three texts, none of
    Message 1 of 4 , Aug 8, 1997
    • 0 Attachment
      Dear listers:
      A textual stemma has a general form that may need to be determined
      before the stemma is decided upon. For example, with three texts,
      none of which can be intermediary between the others, the following
      stemmas are possible: any one of the three may be the archetype,
      having a lost text as a descendant which is the common ancestor of
      the other two texts; or the archetype may be the common ancestor of
      any one of the three texts and of a lost text which again is the
      common ancestor of the other two texts; or the archetype may be a
      lost text, the common ancestor of the three extant texts -- seven
      possibilities, all having a single abstract form: each text connected
      independently to a lost text, A to lost, B to lost, and C to lost.
      An archetype has all the correct readings in the texts, and if no
      one text has all of them, we need to recognize the abstract form to
      decide among the other possibilities and to reconstruct the archetype
      (with the three texts in the example, when they agree, the lost text
      agrees with them; otherwise it agrees with any two of the extant
      texts). This is quite a different procedure from choosing the
      text with the most best readings and substituting the rest in it,
      leaving the great preponderance of its readings unweighed in the
      balance.
      With more than three texts, the stemma may have one or more
      "rings" in it if nothing is done to remove them. If A and B have
      "yes" where C and D have "yea" and A and C have "no" where B and
      D have "nay" then the abstract form of the ring is A--B ("yes"), B--D
      ("nay"), D--C ("yea"), and C--A ("no"). The archetype may be any of
      the four texts or the common ancestor of any two that are connected,
      say, of A and C, making B a descendant of A and D the common
      descendant of B and C (i.e., when A and C disagreed, sometimes A and
      sometimes C had the best reading; the archetype would then read "no"
      whether or not that was one of the identifiably best readings).
      To "break" such a ring we decide which is the "weakest link," all
      variations considered, and before we look for the archetype -- note,
      before we look for the archetype. If, for example, there are over all
      fewer AB agreements than AC agreements or BD agreements or CD
      agreements, then we say that the agreement of A and B in having "yes"
      is abnormal (without deciding whether it is accidental, coincidental
      or emendatory) and set it aside. In effect, then, we have three
      variant readings, "yes (A)," "yes (B)" and "yea," and the abstract
      relationship among the four texts is A--C--D--B. Other evidence may
      modify the relationship without changing the basic series: a lost
      text may take the place of C with C connected to it, or a lost text
      may take the place of D with D connected to it, or there may be two
      lost texts, with C connected to one and D to the other, and one of
      these lost texts may be the archetype.
      Now, using A--B--D--C--A for the ring we have just been
      considering, there may also be rings A--B--C--D--A and A--
      C--B--D--A, three rings that we may call alternates. And if there are
      additional texts we may have two small rings A--B--D--C--A and
      A--B--E--F--A within a larger ring A--C--D--B--E--F--A, the two
      smaller of which we may call connected.
      In order to be consistent in breaking the rings and to divide as
      few links as possible, we need to break them in a certain order and
      may need to repeat the process after the first round of breaking.
      These increasingly complicated matters are treated on pp. 93-98 of my
      book Principles and Practice of Textual Analysis. If those who wish
      to take up my challenge wish also to have a Xerox copy of these pages
      I shall be happy to send them one. So what is my challenge?
      You may decide that you can live with a few rings in your stemmas.
      You will then find that with a medium sized New Testament book like
      First John and the twenty earliest surviving texts, there are nearly
      7000 rings. I have written a computer program that patiently
      identifies all the smaller rings, breaks the rings in the required
      order, and then repeats the process as often as necessary. This
      program traces many linkages that do not lead to rings, and if, for
      example, A leads to C, C to D, D to B, B to E, E to F, and F does
      not lead to A, the program has (I fear) wasted some time. It is smart
      enough to know that if A leads to B but never to any other text
      to the exclusion of B then A--B will never be a link in any ring. It
      might be smartened up so as to know that if F does not lead to A, as
      in the longer example above, then it may be that F will never lead to
      A and the search can stop short when E leads to F. But I have a
      feeling that the initial form of a stemma is like a great net and
      that you ought to be able to move forward from any place or tag end
      on its circumference breaking the meshes (rings) as you come to them
      until the remaining strands spread out like fans within fans, not a
      cross link among them. Do you have or can you divise an algorithm for
      doing this? You do not need to be a computer programmer -- a step by
      step explanation will do, showing how to start with one text and comb
      out the linkages between it and the rest. Remember, unless one of the
      texts has all the best readings it is impossible to decide on the
      archetype until the comb-out is completed.
      This is such a complex matter that it may be better for us to
      communicate directly hereafter instead of through tc-list, but a good
      idea deserves wide recognition, so suit yourselves.
      Vinton Dearing
    • Robert B. Waltz
      On Fri, 8 Aug 1997, Vinton A. Dearing wrote: [ BTW -- I know Dearing asked us to respond off-list. But I have questions that no one
      Message 2 of 4 , Aug 9, 1997
      • 0 Attachment
        On Fri, 8 Aug 1997, "Vinton A. Dearing" <dearing@...> wrote:

        [ BTW -- I know Dearing asked us to respond off-list. But I have
        questions that no one else seems to be asking, so I thought I would
        post here. ]

        >Dear listers:
        > A textual stemma has a general form that may need to be determined
        >before the stemma is decided upon. For example, with three texts,
        >none of which can be intermediary between the others,

        A minor point: Should this not be *directly* intermediary? It is
        usually easy to show that A is the parent of B. It is much harder
        to show that A is the ancestor of B with intervening mixture. I
        assume, from the comments in the rest of this paragraph, that you
        are simply saying that none of the manuscripts are parents of any
        of the others. Correct?

        [ ... ]

        > An archetype has all the correct readings in the texts, and if no
        >one text has all of them, we need to recognize the abstract form to
        >decide among the other possibilities and to reconstruct the archetype
        >(with the three texts in the example, when they agree, the lost text
        >agrees with them; otherwise it agrees with any two of the extant
        >texts). This is quite a different procedure from choosing the
        >text with the most best readings and substituting the rest in it,
        >leaving the great preponderance of its readings unweighed in the
        >balance.

        While I generally agree with the reconstruction method proposed
        (except that I would substitute "text-types" for "texts"), I am
        not sure I understand your distinction. In the final sentence, by
        "best readings," do you mean based on internal evidence or on
        majority rule? For, if the latter, I do not see the distinction
        between the two methods.

        > With more than three texts, the stemma may have one or more
        >"rings" in it if nothing is done to remove them. If A and B have
        >"yes" where C and D have "yea" and A and C have "no" where B and
        >D have "nay" then the abstract form of the ring is A--B ("yes"), B--D
        >("nay"), D--C ("yea"), and C--A ("no"). The archetype may be any of
        >the four texts or the common ancestor of any two that are connected,
        >say, of A and C, making B a descendant of A and D the common
        >descendant of B and C (i.e., when A and C disagreed, sometimes A and
        >sometimes C had the best reading; the archetype would then read "no"
        >whether or not that was one of the identifiably best readings).

        While I concede the possibility of a "ring" in a point of variation
        involving four texts and two points of variation, I do not think
        it possible to create such a grouping if one includes more manuscripts
        and more readings. For example, with four manuscripts and two binary
        readings, there are sixteen possible breakdowns of results. Adding
        just one more reading doubles this. Adding a fifth manuscript increases
        the possibilities by 25%. And so on. The only way one can find "rings"
        is to confine one's self to very small samples of the text. But if
        one is so confined, how does one decide, of the three variants
        yes/yea, no/nay, and should/shall, whether to focus our attention
        on yes/yea and no/nay to the exclusion of should/shall?

        > To "break" such a ring we decide which is the "weakest link," all
        >variations considered, and before we look for the archetype -- note,
        >before we look for the archetype. If, for example, there are over all
        >fewer AB agreements than AC agreements or BD agreements or CD
        >agreements, then we say that the agreement of A and B in having "yes"
        >is abnormal (without deciding whether it is accidental, coincidental
        >or emendatory) and set it aside. In effect, then, we have three
        >variant readings, "yes (A)," "yes (B)" and "yea," and the abstract
        >relationship among the four texts is A--C--D--B. Other evidence may
        >modify the relationship without changing the basic series: a lost
        >text may take the place of C with C connected to it, or a lost text
        >may take the place of D with D connected to it, or there may be two
        >lost texts, with C connected to one and D to the other, and one of
        >these lost texts may be the archetype.

        Again, I agree that we should determine everything possible about our
        manuscripts before we look for the archetype. But I fail to see the
        point of "breaking the ring." The only justification I can see is to
        cast off one of the four manuscripts so that one can make a decision
        in the event of a two-versus-two tie. (Which, BTW, is *not* my method;
        in a two-versus-two tie I would actually look to internal evidence.)
        But it would appear that your method rewards texts which agree often.
        In the case of a tie, I would be inclined, in the abstract, to reward
        those which did *not* agree often.


        > Now, using A--B--D--C--A for the ring we have just been
        >considering, there may also be rings A--B--C--D--A and A--
        >C--B--D--A, three rings that we may call alternates. And if there are
        >additional texts we may have two small rings A--B--D--C--A and
        >A--B--E--F--A within a larger ring A--C--D--B--E--F--A, the two
        >smaller of which we may call connected.

        Technical footnote: There are actually *six* possible rings with
        four members (assuming we always start with A):

        A-B-C-D
        A-B-D-C
        A-C-B-D
        A-C-D-B
        A-D-B-C
        A-D-C-B

        If you have a method for restricting this to the rings you listed,
        I failed to understand it.

        > In order to be consistent in breaking the rings and to divide as
        >few links as possible, we need to break them in a certain order and
        >may need to repeat the process after the first round of breaking.
        >These increasingly complicated matters are treated on pp. 93-98 of my
        >book Principles and Practice of Textual Analysis. If those who wish
        >to take up my challenge wish also to have a Xerox copy of these pages
        >I shall be happy to send them one. So what is my challenge?
        > You may decide that you can live with a few rings in your stemmas.
        >You will then find that with a medium sized New Testament book like
        >First John and the twenty earliest surviving texts, there are nearly
        >7000 rings. I have written a computer program that patiently
        >identifies all the smaller rings, breaks the rings in the required
        >order, and then repeats the process as often as necessary. This
        >program traces many linkages that do not lead to rings, and if, for
        >example, A leads to C, C to D, D to B, B to E, E to F, and F does
        >not lead to A, the program has (I fear) wasted some time. It is smart
        >enough to know that if A leads to B but never to any other text
        >to the exclusion of B then A--B will never be a link in any ring. It
        >might be smartened up so as to know that if F does not lead to A, as
        >in the longer example above, then it may be that F will never lead to
        >A and the search can stop short when E leads to F. But I have a
        >feeling that the initial form of a stemma is like a great net and
        >that you ought to be able to move forward from any place or tag end
        >on its circumference breaking the meshes (rings) as you come to them
        >until the remaining strands spread out like fans within fans, not a
        >cross link among them. Do you have or can you divise an algorithm for
        >doing this? You do not need to be a computer programmer -- a step by
        >step explanation will do, showing how to start with one text and comb
        >out the linkages between it and the rest. Remember, unless one of the
        >texts has all the best readings it is impossible to decide on the
        >archetype until the comb-out is completed.

        Speaking as a computer programmer, I don't think you've defined the
        problem clearly enough for me to comment on a solution algorithm.
        I don't doubt that you have actually refined your method further
        than what you have described here, but based on the comments you've
        made, I can't add anything.

        -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-

        Robert B. Waltz
        waltzmn@...

        Want more loudmouthed opinions about textual criticism?
        Try my web page: http://www.skypoint.com/~waltzmn
        (A site inspired by the Encyclopedia of NT Textual Criticism)
      • James R. Adair
        Several years ago I read both of Dearing s books and found them both fascinating and informative. He has of course only given us a brief synopsis of his views
        Message 3 of 4 , Aug 10, 1997
        • 0 Attachment
          Several years ago I read both of Dearing's books and found them both
          fascinating and informative. He has of course only given us a brief
          synopsis of his views on the list. Since it's been a while since I read
          them, I want to test my memory a bit and reply to both Vinton Dearing and
          Bob Waltz.

          On Sat, 9 Aug 1997, Robert B. Waltz wrote:

          > On Fri, 8 Aug 1997, "Vinton A. Dearing" <dearing@...> wrote:
          >
          > >Dear listers:
          > > A textual stemma has a general form that may need to be determined
          > >before the stemma is decided upon. For example, with three texts,
          > >none of which can be intermediary between the others,
          >
          > A minor point: Should this not be *directly* intermediary? It is
          > usually easy to show that A is the parent of B. It is much harder
          > to show that A is the ancestor of B with intervening mixture. I
          > assume, from the comments in the rest of this paragraph, that you
          > are simply saying that none of the manuscripts are parents of any
          > of the others. Correct?

          There is no necessity to say "directly intermediary," since with any three
          mss with any given set of readings, even if there is an apparent ring, it
          can be shown that by creating a hypothetical intermediary ms that has the
          majority reading whenever the mss split 2-1, this ms can be added to the
          other three in such a way as to eliminate the ring:

          A A
          / \ --> |
          B---C x
          / \
          B C


          In case not everyone is clear on the definition of a ring, let me try to
          explain it. Dearing's genealogical method begins with the assumption that
          every ms under consideration had only one exemplar, so if a reading is
          present in two mss, they should be genetically related (of course, common
          scribal mistakes such as the addition or omission of a conjunction must be
          eliminated from consideration). From time to time, a scribe might create
          a variant that also appears in an unrelated ms, thus giving the appearance
          of a relationship that is not real. And of course, in the case of
          biblical mss, since we know that sometimes one ms was corrected from
          another, variants from (mostly) unrelated mss can appear in another ms.
          These agreements based on either accident or correction are called rings.

          > > With more than three texts, the stemma may have one or more
          > >"rings" in it if nothing is done to remove them. If A and B have
          > >"yes" where C and D have "yea" and A and C have "no" where B and
          > >D have "nay" then the abstract form of the ring is A--B ("yes"), B--D
          > >("nay"), D--C ("yea"), and C--A ("no"). The archetype may be any of
          > >the four texts or the common ancestor of any two that are connected,
          > >say, of A and C, making B a descendant of A and D the common
          > >descendant of B and C (i.e., when A and C disagreed, sometimes A and
          > >sometimes C had the best reading; the archetype would then read "no"
          > >whether or not that was one of the identifiably best readings).
          >
          > While I concede the possibility of a "ring" in a point of variation
          > involving four texts and two points of variation, I do not think
          > it possible to create such a grouping if one includes more manuscripts
          > and more readings. For example, with four manuscripts and two binary
          > readings, there are sixteen possible breakdowns of results. Adding
          > just one more reading doubles this. Adding a fifth manuscript increases
          > the possibilities by 25%. And so on. The only way one can find "rings"
          > is to confine one's self to very small samples of the text. But if
          > one is so confined, how does one decide, of the three variants
          > yes/yea, no/nay, and should/shall, whether to focus our attention
          > on yes/yea and no/nay to the exclusion of should/shall?

          Bob is right about adding mss and readings. With many mss and many
          readings, one quickly gets not just simple rings but whole networks of
          mss. To break these networks into a simple tree, Dearing proposes
          breaking individual links, starting with the weakest ones, until a simple
          tree is created.

          > > To "break" such a ring we decide which is the "weakest link," all
          > >variations considered, and before we look for the archetype -- note,
          > >before we look for the archetype. If, for example, there are over all
          > >fewer AB agreements than AC agreements or BD agreements or CD
          > >agreements, then we say that the agreement of A and B in having "yes"
          > >is abnormal (without deciding whether it is accidental, coincidental
          > >or emendatory) and set it aside. In effect, then, we have three
          > >variant readings, "yes (A)," "yes (B)" and "yea," and the abstract
          > >relationship among the four texts is A--C--D--B. Other evidence may
          > >modify the relationship without changing the basic series: a lost
          > >text may take the place of C with C connected to it, or a lost text
          > >may take the place of D with D connected to it, or there may be two
          > >lost texts, with C connected to one and D to the other, and one of
          > >these lost texts may be the archetype.
          >
          > Again, I agree that we should determine everything possible about our
          > manuscripts before we look for the archetype. But I fail to see the
          > point of "breaking the ring." The only justification I can see is to
          > cast off one of the four manuscripts so that one can make a decision
          > in the event of a two-versus-two tie. (Which, BTW, is *not* my method;
          > in a two-versus-two tie I would actually look to internal evidence.)
          > But it would appear that your method rewards texts which agree often.
          > In the case of a tie, I would be inclined, in the abstract, to reward
          > those which did *not* agree often.

          In the case of a tie, as I recall, Dearing's method says that the it
          doesn't matter which link is broken--either is equally likely to be the
          correct way to break it. I think it would interesting to combine this
          purely mechanical approach with a subjective evaluation of internal
          readings and see what happens.

          > > Now, using A--B--D--C--A for the ring we have just been
          > >considering, there may also be rings A--B--C--D--A and A--
          > >C--B--D--A, three rings that we may call alternates. And if there are
          > >additional texts we may have two small rings A--B--D--C--A and
          > >A--B--E--F--A within a larger ring A--C--D--B--E--F--A, the two
          > >smaller of which we may call connected.
          >
          > Technical footnote: There are actually *six* possible rings with
          > four members (assuming we always start with A):
          >
          > A-B-C-D
          > A-B-D-C
          > A-C-B-D
          > A-C-D-B
          > A-D-B-C
          > A-D-C-B
          >
          > If you have a method for restricting this to the rings you listed,
          > I failed to understand it.

          There are only three possible rings. Remember that since the mss are a
          ring, the last is connected to the first. Thus, in Bob's list above, the
          last three groups of mss are equivalent to the first three (#4 = #2, #5 =
          #3, #6 = #1). Maybe it's clearer to draw it this way:

          A-B A-C A-C
          | | | | | |
          D-C B-D D-B

          > > In order to be consistent in breaking the rings and to divide as
          > >few links as possible, we need to break them in a certain order and
          > >may need to repeat the process after the first round of breaking.
          > >These increasingly complicated matters are treated on pp. 93-98 of my
          > >book Principles and Practice of Textual Analysis. If those who wish
          > >to take up my challenge wish also to have a Xerox copy of these pages
          > >I shall be happy to send them one. So what is my challenge?
          > > You may decide that you can live with a few rings in your stemmas.
          > >You will then find that with a medium sized New Testament book like
          > >First John and the twenty earliest surviving texts, there are nearly
          > >7000 rings. I have written a computer program that patiently
          > >identifies all the smaller rings, breaks the rings in the required
          > >order, and then repeats the process as often as necessary. This
          > >program traces many linkages that do not lead to rings, and if, for
          > >example, A leads to C, C to D, D to B, B to E, E to F, and F does
          > >not lead to A, the program has (I fear) wasted some time. It is smart
          > >enough to know that if A leads to B but never to any other text
          > >to the exclusion of B then A--B will never be a link in any ring. It
          > >might be smartened up so as to know that if F does not lead to A, as
          > >in the longer example above, then it may be that F will never lead to
          > >A and the search can stop short when E leads to F. But I have a
          > >feeling that the initial form of a stemma is like a great net and
          > >that you ought to be able to move forward from any place or tag end
          > >on its circumference breaking the meshes (rings) as you come to them
          > >until the remaining strands spread out like fans within fans, not a
          > >cross link among them. Do you have or can you divise an algorithm for
          > >doing this? You do not need to be a computer programmer -- a step by
          > >step explanation will do, showing how to start with one text and comb
          > >out the linkages between it and the rest. Remember, unless one of the
          > >texts has all the best readings it is impossible to decide on the
          > >archetype until the comb-out is completed.

          Long ago I wrote a program myself (using a method somewhat different from
          Dearing's) that would create a tree from a given set of mss and readings,
          but it didn't break the rings, only identified them. My question in
          regard to the whole approach is this. Is it valid to use the model of a
          simple tree when dealing with the transmission of the biblical text? This
          question can't be answered in a strictly theoretical way--data is needed.
          At what point in the transmission of the NT, for example, did scribes
          begin correcting their newly copied mss from a second exemplar? Harry
          Gamble, in _Books and Readers in the Early Church_, seems to imply that
          the majority of early mss were created privately by copying a single
          exemplar. When did the practice of correcting the text against a second
          ms become widespread enough so that, after that point, the idea of using a
          simple tree to model the transmission of the text breaks down?

          Jimmy Adair
          Manager of Information Technology Services, Scholars Press
          and
          Managing Editor of TELA, the Scholars Press World Wide Web Site
          ---------------> http://scholar.cc.emory.edu <-----------------
        • Vinton A. Dearing
          I am grateful to Bob Waltz for his second communication. If there are others who wish a more detailed explanation of the problem as I see it and who wish to
          Message 4 of 4 , Aug 12, 1997
          • 0 Attachment
            I am grateful to Bob Waltz for his second communication. If there are
            others who wish a more detailed explanation of the problem as I see
            it and who wish to tackle it, please let me know.
            Vinton Dearing
          Your message has been successfully submitted and would be delivered to recipients shortly.