mls@sfsup.att.com (Mike Siemon) (03/14/91)
Well, I think I gave O'Keefe too much credit. I have looked at his "test"
and find it abysmally inadequate to its stated purpose. I am not about to
mail him a damn thing, but I *will* post my analysis here on March 21. For
now, I will make just a few generic comments.
(1) he seems to think that 10 out of his 14 sentences (71+% correct) is the
appropriate marker of a successful response. I wonder where he gets
this number? Why is THAT the marker of a "good" answer, and not 60% or
80%? The null hypothesis (no recognizable distinction) would be 50%,
or so I would assume unless a better experimentalist corrects me. As I
said in my overall critique of the experimental design, we have NO basis
for specifying what would be a significant difference from the null hypo-
thesis.
(2) O'Keefe gives 14 sentences, containing by my count (or by UNIX's count :-))
329 words. This is (as even he acknowledges) an excruciatingly small
sample. By way of contrast, the Noah story in Genesis occupies some 68
verses and 1817 words. This isn't *quite* an order of magnitude differ-
ence, but it's pretty major.
(3) If equally divided between his two sources, his passage has 7 sentences
from each -- if not, then one of them is represented by even fewer than
7 sentences. I do not see how ANY serious critical examination can be
built on so small a sample.
(4) O'Keefe offers as one of the "distinguishing" features of the two authors
that one is male and the other female. I am unaware that that makes any
stylistic difference -- especially in academic prose which at least the
female "medical historian" may be presumed to be writing.
(5) Both are writing at the SAME date (early 1980's), about highly documented
episodes that happened less than a century before.
(6) There may be a stylistic difference in that one author is American and
the other British. That is potentially helpful -- but it is amusing to
note that NONE of the characteristic differences (as perennially posted
on soc.culture.british or other usenet groups) between these dialects
is particularly obvious in the body tested (there ARE American colloquial-
isms in the "sample" of the American author.) This may signify nothing
more than the avoidance of dialect in general "intellectual" discourse;
but if there ARE dialect differences in the OT, this choice misrepresents
the data it attempts to make a comparison to.
(7) The samples of the two authors show a number of significant stylistic
differences; it is unclear from the way O'Keefe poses his challenge if
these samples are indeed from the SAME texts -- or ones that we may in
confidence take to be representative -- of the sources for his selection.
I will, on March 21, give my list of what these style markers are, and
how they may apply to the test sentences. Let me say up front that the
style markers for author A are *not* much in evidence in the test text
-- and one reason for that may be that the "sample" for author A is almost
entirely an indirect quote of someone ELSE (the clinician Abe). One may
reasonably wonder if the style markers we are given are those of A or of
Abe.
In short, I think that O'Keefe has offered a totally misconceived test case
that hardly addresses the issues the D.H. supporters claim to find and resolve.
I think that I will also post, on March 21, my own conflation test. No doubt,
as I said before, my biases will generate an "easier" test than O'Keefe has
done -- nonetheless, I hope to amuse and instruct you in my effort.
--
Michael L. Siemon "O stand, stand at the window,
m.siemon@ATT.COM As the tears scald and start;
...!att!attunix!mls You shall love your crooked neighbor
standard disclaimer With your crooked heart."