From: DWILKINS@ucrac1.ucr.edu
Date: Fri Jul 26 1996 - 14:11:06 EDT
Erik van Halsema said:
>For instance: suppose that in 1 Cor. 8-10 the average sentence length is 15
>(bypassing the problem: 'what is a sentence'??...).
I've been away from e-mail a day or two and I just can't seem to resist the
temptation to comment before looking through all the posts first, so my usual
apology for redundancy if someone has already said this. Erik, I haven't had
formal statistics training but have done some crash study on the subject in
relation to the very thing you're talking about. In fact, I am writing a com-
puter program which, inter alia, checks and averages sentence length in Greek
and Latin texts (hopefully I will soon be making an alpha version available
to list members--it reads and searches the TLG CD). The biggest problem in
such a study is "what is a sentence?", so unfortunately you can't bypass it.
If you simply look for periods and question marks, your results will be next
to worthless. You have to take into consideration that the original writers
did not use punctuation, but tended to connect all their sentences with various
conjunctions. Indeed, in the NASB '95 Update (which is just recently being
published) we took a major step and eliminated many of the conjunctions at
the beginning of "sentences" (i.e. what are punctuated as such in modern
editions of the Greek NT) largely for this reason. Studies have been done in
the past on average sentence length in various works, and have been duly
criticized for overlooking this basic problem. Paul is especially hard to
evaluate on this level because he seems to excell in tying his sentences to-
gether with conjunctions. Not that such a statistical evaluation would not
be useful in identifying stylistic tendencies, but the first step is to do a
convincing job of identifying true sentences. Using the superficial method
of searching end punctuation would have to have a rather high standard devi-
ation, I think. BTW, I believe that the magic number in stats is 30, which in
this situation probably means that you have to evaluate at least 30 comparable
authors to determine what the average length would be, and perhaps 30 separ-
ate works by one author to determine his/her average length (this is a point
about which I am very ignorant, and hope that someone who knows statistics
well can-or already has-given us clarification and reason to believe that
such a task is mathematically possible, given so little data to work with).
Don Wilkins
UC Riverside
This archive was generated by hypermail 2.1.4 : Sat Apr 20 2002 - 15:37:46 EDT