Response to Jockers, Criddle, et al., Now Available

[_Roger]

Chris:

Unless you can point out a flaw in the design of Bruce's study, that is what it shows.

I believe the flaw lies in the word-frequency method itself (at least as it has been applied to the Book of Mormon).

But if you accept the validity of the Jockers method, then you'll have to either find some other flaw in Bruce's experimental design, or accept his conclusions.

That would be something like an armchair quarterback telling Adrian Rogers where his passing technique is off.

Nevertheless as silly as that scenario would be, I am inclined to think there is a flaw in there somewhere. I suspect it centers around the fact that whoever wrote the Book of Mormon was attempting to imitate the language of the Bible but the 19th century samples were not.

As a layman, that's about as far as I can go at this point.

[_GlennThigpen]

As a layman, that's about as far as I can go at this point.

As a layman you can note that the flaw you suspect in Bruce's work would also apply to the Jockers study also.

However, if that were the case, you would expect that Isaiah would show up as the author of a lot more of the chapters. There are a few false positives for Isaiah outside of 1 and 2 Nephi where he is quoted extensively but no more so than the other candidates.



Glenn

[_Uncle Dale]

...
As a layman you can note that the flaw you suspect in Bruce's work would also apply to the Jockers study also.
...

Except for Bruce's initial step -- which is to produce a pca chart
in which 19th century authors' word-printed writings plot far
away from Book of Mormon chapters. Bruce places great value
in that particular graphic -- as providing solid evidence that those
19th century authors are not even eligible for NSC analysis, because
their writings are not anything like those found in the Book of Mormon.

Jockers and associates also produced a set of statistical charts in
their preliminary screening of author-candidates, and were convinced
that a somewhat similar plots configuration on one of their own
charts was NOT a determining filter for eliminating author-candidates.

So, we start out with that basic difference, before any determinative
computer analysis is even begun.

For Jockers and associates, the fact that the Book of Mormon biblical chapters
cluster with the Book of Mormon non-biblical chapters presents a probable
solution as to why the 19th century writers' works do not greatly
overlap the "Book of Mormon cloud" of texts in pca chart no. 1.

Obviously Nephites did not write Isaiah, Malachi and Matthew -- so
the clustering of non-biblical Book of Mormon chapters around those replicas
of biblical texts must occur for reasons OTHER than authorship.

Because the pca chart #1 clustering of Book of Mormon chapters, along with
the Book of Mormon biblical texts, is not definitely due to authorship, then
the separated 19th century texts on the same chart cannot be
ruled out of planned testing, due to "authorship."

As it turned out, Bruce did not rule them out of testing in his own
study -- and, in fact, added one more -- Joseph Smith.

Smith is a viable author-candidate (no matter pca charts) because
we are told that he did indeed write part of the Book of Mormon (the Preface).

So -- before the Delta and NSC testing was ever conducted (for
either study) Bruce and Matt started out from differing conclusions
regarding viable author-candidates.

AFTER that initial step, you might say that "the flaw you suspect
in Bruce's work would also apply to the Jockers study also," with
some special relevance -- because neither group of researchers
took the trouble to calculate the degree of impact upon wordprints
that conscious emulation of archaic English might produce.

That is the first "flaw," as I see it.

Hopefully somebody will eventually produce a similar computerized
wordprint study in which emulation of archaic English is measured
and considered as a factor in altering authors' wordprints.

UD

[_GlennThigpen]

Hopefully somebody will eventually produce a similar computerized
wordprint study in which emulation of archaic English is measured
and considered as a factor in altering authors' wordprints.

UD

There have been studies that have concluded that a person's non contextual word pattern usage does not change even when trying to mimic another author's style. An example is in A. Q. Morton's "Literary Detection" where an author had tried to mimic Jane Austen's literary style in finishing a book started but not completed before her death in 1817. The mimicry was good, but the non contextual word pattern usage gave her away.
However, this level of uncertainty concerning the effectiveness of the NSC method and genre leave the Hilton study as the best one to date.
It was not deemed sufficient by the Jockers team because of the size of the texts needed for reliable results and thus could not be used for a chapter by chapter comparison. I have never been convinced that this was a particularly strong point against the Berkeley Group's findings. No one has produced any type of analysis to warrant a suspicion that the books such as 1 and 2 Nephi, Mosiah, Ether, Alma, etc have multiple authors except where the text indicates such.
We have a portion of the original manuscript and know how it was written, and the printer's manuscript as evidence that it would have been almost impossible for any major collaboration to have taken place. There are numerous conspiracy stories that have nebulous foundations for the edification of those who do not accept the miracle story.
But Dale is right. We need to find a lot more author candidates, since the earlier word print studies identified 24 authors with their unique word print styles. Amazingly enough, these authors checked off against each other very well. That is 1 Nephi and 2 Nephi, excluding the Jacob parts and Isaiah quotes matched while they differed from Alma, Mosiah, etc. If those chapters were a mish mash of collaborators, one would expect that they would not match each other so well.

Glenn

[_CaliforniaKid]

The earlier study to which you refer was seriously problematic.

[_Uncle Dale]

...There have been studies that have concluded that a person's non contextual word pattern usage does not change even when trying to mimic another author's style.
...

That may be, if we are merely talking about "style."

But, if we are talking about significantly altering grammar and
vocabulary, then that's a whole new ball game.

If I had two stories you wrote -- one in your "regular voice,"
and one in which you tried to emulate Rush Limbaugh, then I
think that wordprinting would say you were the author of both.

However -- if I took one of those two texts, and inserted
"and it came to pass," "now it came to pass," "but behold,"
"and now behold," etc., hundreds of times, I'm sure I could
so alter the text as to make it mis-match your wordprint.

The sorts of words used in wordprinting are "now," "it," "and,"
"to," "but," etc. -- possibly even "came" and "pass."

If I added many, many new occurrences of those words to
one of your texts, I could probably mask your regular wordprint
significantly ---- However, I probably would not thus change
the ratios of ALL your frequently occurring non-contextual words.

In some ways, the altered word print would still match up with
your other writings.

But -----------> if I asked a computer program to seek out the
ways in which the altered text LEAST matched with your other
writings, then I would obviously be asking it to find a way to
SEPARATE the altered textual segments from your regular writings
on any scatter diagram chart I wanted to generate.

Get my point?

UD

[_GlennThigpen]

Get my point?

UD

Yes, and I agree with you. However, if filters for those phrases such as "and it came to pass" were established, that would eliminate that skew factor. I'll have to reread the Larsen, Wrencher, layto paper again, but I seem to recall that they were aware that such phrases would skew the results, as were the Berkeley Group.

Glenn

[_Benjamin McGuire]

MCB writes:

The body of literature you are looking at is different from the one I am looking at.

And earlier, you wrote:

The first half is a paltry patch-up job.

That I may be using different sources than you doesn't change the fact that I can make a good argument that the first half of the Book of Mormon is not a "paltry patch-up job" but is in fact a fairly consistent and complex piece of literature that has a great deal of coherence in its narrative.

Ben McGuire

[_Uncle Dale]

Get my point?

UD

Yes, and I agree with you. However, if filters for those phrases such as "and it came to pass" were established, that would eliminate that skew ....

Glenn

Such manipulations would have to be done intelligently and
very, very carefully -- in order to produce useful results.

Emulating archaic English does not just introduce extra
non-contextual words as a sort of surplus -- it also replaces
some of the words from that group that might be expected
in a "normal" text with its own favored expressions.

But there is also the effect of grammar -- one example of
which is the emphatic case in verb conjugations.

Modern English = I went

Archaic English = I did go; you do go; he doth go; thou goest

Thus, if "went" were included in the package of non-contextual
words making up a word-print, the application of archaic English
emulation might totally screw up its occurrence ratio -- along
with multiple other non-contextual words affected by grammar.

I'm not saying that it would be impossible to strip Book of Mormon chapters
of the worst "offenders" that skew non-contextual word counts;
but it would be tricky and controversial, to say the least.

Were such a "correction" methodology agreed upon, and applied
in NSC authorship attribution, then I'm fairly confident that the
Book of Mormon non-biblical chapters would cluster less closely
with that book's biblical reproductions, in Bruce's pca charts.

Would such an adjustment bring the 19th century authors' plots
any closer to the Book of Mormon chapters' plots on that same chart?

I'm unsure of the results -- but in looking back at the pc2 and pc3
charts in that same Jockers-data series, I suspect that the overlap
we see there (of Book of Mormon and non-BoM texts) might carry over
somewhat into a "corrected" pc1 scatter diagram.

UD

[_CaliforniaKid]

Emulating archaic English does not just introduce extra
non-contextual words as a sort of surplus -- it also replaces
some of the words from that group that might be expected
in a "normal" text with its own favored expressions.

Yes, very well said. I think you've put your finger on the pulse of the problem for any stylometric method relying on common or "non-contextual" words.