Wed May 1 14:34:02 1996

From majordom  Wed May  1 14:34:02 1996
Return-Path: 
Received: by scholar.cc.emory.edu (5.0/SMI-SVR4)
	id AA17467; Wed, 1 May 1996 14:34:02 +0500
Date: Wed, 1 May 96 13:30 CDT
X-Sender: waltzmn@popmail.skypoint.com
Message-Id: 
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: tc-list@scholar.cc.emory.edu
From: waltzmn@skypoint.com (Robert B. Waltz)
Subject: Mathematical Methods... A Modest Proposal :-)
Content-Length: 2835
Sender: owner-tc-list@scholar.cc.emory.edu
Precedence: bulk
Reply-To: tc-list@scholar.cc.emory.edu

Since I don't want this to descend into an all-mathematically-inclined-people
war, may I offer a proposal:

Can we, perhaps, come up with an agreement on what, exactly, we think is
needed to accurately compare manuscripts? If we can state what we would
ideally need, maybe we can decide how close to this we can come.

The criteria I would look for (and note that I did _not_ claim to
achieve them all, merely to have come closer than previous attempts)
would include the following:

1. Large sample sizes (at least a thousand readings in the gospels,
   300 in Acts, 750 in Paul, 200 in the Catholics, 250 in the Apocalypse).
   This not only increases the polling accuracy of our sample, it allows
   us to examine individual books and parts of books IF we need to.
   It also gives us a better shot at classifying fragments.

2. Unbiased but intelligent selection of readings. By "intelligent"
   I mean that we don't use too many variants of the type "add/omit
   article," or the like, which cannot be rendered in the versions,
   nor too many variants where scribes could easily have made a
   change my accident (e.g. between future indicative and aorist
   subjunctive). By "unbiased" I mean that we not take many variants
   of the "D-and-a-few-Old-Latin-against-the-world" type.

3. A mathematical methodology that counts more than simple agreements.
   For example, I have worked with overall agreements (which seek
   for sister manuscripts), agreements where both manuscripts are
   non-Byzantine (for seeking related mixed manuscripts), agreements
   where the Byzantine text is divided (for classifying Byzantine
   groups), and "near-singular readings" -- readings which are
   sufficiently unusual that they probably represent derivation from
   a common ancestor.

4. A complete independence from previous results. We must not start with
   preconceptions, and we must not assume it in our results. One of the
   worst mistakes I ever made was in assuming that the lists of
   "Alexandrian" manuscripts in Metzger and Greenlee were correct. I
   trusted them, and created a definition of the "Alexandrian" text in
   my database. And needed most of a year to recover from the mistake.

I retain the right to add to this list. This is just what sprang to mind
instantly.

I'd also like to know what people are looking for in such results. Colwell
gave offered a proposal -- the famous (or infamous) 70/10 rule --
and eventually fathered the Claremont method. Claremont works for sisters,
but can't handle mixed manuscripts -- and also was biased by previous
assumptions. The 70/10 rule works only if one studies un-mixed manuscripts.
We need something better.

We probably cannot, at this stage, achieve ALL our goals. But until we
state them, we cannot decide how close we can come.

Bob Waltz
waltzmn@skypoint.com

Back