DCHindley

You might check out Anthony Kenney (I hope I didn't get that wrong, my library is all in boxes in the basement fort a while longer). A Q Morton co-wrote a study entitled Paul: the Man and the Myth (or via: amazon.co.uk) which studied mainly sentence length and to a lesser degree use of various grammatical features.

I once actually input his data into a spreadsheet and created graphs of the sentence length distributions for each of the letters of Paul, the general epistles (James, etc), and a few early fathers (Clement of Rome and maybe a couple others).

You are quite correct that it is possible to use stylostatistical analysis when you have a sizeable, unquestionably authentic corpus to use as a control against the unknown sample. You also need a suitably large sample to ensure that the results aren't significant simply by chance.

Unfortunately, the individual letters of the Pauline corpus are just too small to yeild high confidence in an analysis, and besides, there is a methodological problem inherent in using the entire corpus as the control for analysis of a specific book within the corpus.

Dave

Toto

Do you mean this Anthony Kenny - author of The Computation of Style: an introduction to statistics for students of literature and humanities ?

Googling it brings up this Masters Thesis, which I have not read yet:

Quantitative Authorship Attribution: A History and an Evaluation of Techniques

=====

More on Morton:

A Hypertextual History of Humanities Computing: The Pioneers

Kosh

I saw something like this once. But when I tried to read the code, it was all greek to me....

Thank you. Thank you very much.

Julian

There is very little material on the topic of authorship and/or interpolation detection. I have written a bunch of software to do just that and I have blogged quite a bit on some of the methodology (see early entries here: http://julian.textcrit.com) that can be used on words. What I didn't have, and what would help immensely, is a syntactically parsed text. I have since been working on creating just that and am quite close. Once we have syntactic information we add a whole new layer of information to our stylometric representation of a text. There are a number of papers on how all these factors help by using google: http://www.google.com/search?hl=en&s...eek+authorship I am not sure how many of these you will be able to read since most are protected, I have excellent access from my office here at the patent office.

The main problem is that any computer methods work less and less well the less text you have. Most of the time an interpolation is so short as to yield little to no information.

Julian

Kosh

Hopefully you have been using something like the UNIX utility "lex" (lexigraphical analyzer) to help you write the parser?

I'm sure there's something better than that out there these days.

Julian

No, that would not be possible. A simple CFG is not sufficient to parse a free word order language like Greek. Lex, yacc and all those things are very nice but insufficient. I did use a CFG (context-free grammar) and Parse::RecDescent to define my compiler compiler (not a mistake, I did mean to write that twice) but there are no simple rules that can define rules for natural language processing. That's why there are so many approaches like LFG, SFG, X-Bar, CDG, etc, etc... My approach is a hierarchial multi-pass linear sequence analyzer combined with a dependency grammar algorithm and that seems to do a pretty decent job.

Julian

DCHindley

Well, actually no.

While Anthony Kenny (not Kenney, my bad) wrote _A Stylometric Study of the New Testament_ (Oxford-New York-Toronto-Sydney-Paris-Frankfurt, 1986), I was actually thinking of Kenneth J. Neumann, The Authenticity of the Pauline Epistles in the Light of Stylostatistical Analysis (or via: amazon.co.uk) (Atlanta, 1990).

I have really got to get my library out of those boxes.

Dave

DCHindley

I think we should distinguish between the confirmation of possible authorship from detection of interpolations. As I mentioned above, I do not believe it is possible to determine whether any of the 13 books attributed to Paul are *actually* his, due to limitations of the materials that have been preserved.

That being said, I do believe that stylostatictical analysis can help us determine the degree to which a text seems to be written by the same author throughout (homogeneity).

Interpolations, at least as I use the term, refers to glosses or other words and phrases that get added to an original author's text by an editor or redactor, not a copyist making spelling or grammatical corrections or accidentally including a marginal note.

One could mark off text that one thinks is intrusive to the sentence or argumentative structure, or offers an explanation, or is a form of commentary, and see if the remaining text is more homogeneous, but that can be a slippery slope. Some authors are more complex writers than others, so reducing the number of aporia or anomolies may be deceptive.

This is the problem I have been grappling with in my ever so evil and obviously wrong hypothesis that the Pauline letters were originally written by a non-christian who felt that gentiles who trusted in the promises made to Abraham by God were justified in God's sight without having to fully convert to Judaism.

These letters were later adopted and adapted by a radically anti-semite editor/redactor who added all sorts of digressions and glosses to make Paul into a good christian of his own era.

Dave Hindley

Clivedurdle

The Spruce Goose flew!

(And why no further mention of Ellegard in this thread?)

Toto

Interesting synchonicity - this article just popped into my inbox:

Shedding Light on The Dark Tower: A C.S. Lewis mystery is solved.

But - a witness has been found who authenticates the work as Lewis'.