Counting Guns: Response to Paul Finkelman
jlindgren at WORLDNET.ATT.NET
Mon Jan 15 00:14:58 PST 2001
As a friend of mine, Paul F. should have checked his facts with me
Paul Finkelman wrote:
> I don't want to prolong this debate, but it is not entirely clear that
> "Bellesliles got the factual record
> wrong and the lawprof (Lindgren) got it right" as Frank Cross
asserts. What is
> clear is that both are looking at different records (as well as some
> overlapping ones) and that the "record" may be confused and muddled.
> has not looked at the manuscript sources; he has not gone through
> of probate records as Bellesiles has (Lindgren asked for the data
> used, not knowing, I think, that Bellesiles had about 100,000 pages
> notes). Lindgren has looked a few printed records and found some
Paul's last point is recklessly false. I have looked at more
than published records. Further, the main data in my manuscript is
EXACTLY the same published records as Bellesiles says he used in Arming
and in his 1996 J. of Am. History article (which published the 1765-90
percents for the
first time). We looked at other records as well, but report that data
Paul F. did not ask me what records I have looked at before he posted,
so how could he know? To present his speculations as fact (once again,
as he did recently for Sandy
Levinson) is really too much. Paul, next time ask me what I read before
posting; I like to
understand my subject as fully as possible before writing (and to be
careful in my claims on such an explosive issue).
In Arming America, Bellesiles never says how many records he looked at.
He took a large number of unrelated samples and gives no counts of those
sample sizes. He gives
only an approximate cell count (sample size) for one cell in his table,
1,200 inventories for the Frontier (1765-90). Nowhere does he give any
other cell counts (sample sizes) for the other 29 cells in his table 1
(AA,p. 445). A Ph.D. historian tells me that Bellesiles told him that
his approximate count for the first column of his table (the years
1765-90) was about 4,000 inventories. Bellesiles has not provided any
on repeated request (so I can't know how many he looked at or analyzed,
whether the 4,000 number is correct), nor has he ever said whether he
does or does
not have the counts (sample sizes) for the data in his tables. I sure
wish he'd actually give the cell counts for each cell so we can know for
sure whether he looked at more
probate records than we did for the period of our disagreement and, if
how many more. Also good would be to release precise cites to his
with counts for each source.
I have looked at about 2,500-3,000 inventories for the 1765-1790 period
300 inventories for earlier periods), not that different than the 4,000
estates Bellesiles reportedly claimed to another researcher. More than
the inventories I examined for 1765-90 were in microfilms of the
original records, the rest in published sources. In addition, my
co-author Justin Heather has looked at other
large runs of unpublished probate records from the same period. In our
paper we provide precise counts for a total of about 1,452 inventories
and cite the records. This more inventories than Bellesiles actually
cites to records for in his article and book combined. Between the two
works, Bellesiles cites as having used only the Jones database and the
Providence database, together totalling about 1,100 inventories. The
rest of his 11,000+ probate inventories remain uncited to this day.
One must remember that Alice Hanson Jones, Anna Hawley, and Harold Gill
have themselves independently counted guns in various databases and
their counts roughly match ours rather than Bellesiles'.
We didn't include the unpublished records for several reasons, not the
least of which that deciphering handwriting and inaccessibility of
records can lead to claims that are difficult to check. Our IMPRESSION
is that the patterns we saw were similar to those in the data we
published, though we didn't carefully count most counties we looked at.
I am uncomfortable even mentioning these data and do not want anyone to
rely on these impressions since we have not done careful enough counts
to give anything more than our impressions. Every county we looked at
APPEARED to have more guns than Bellesiles' averages would suggest for
the applicable region. That we didn't publish these data doesn't mean
we didn't look at them, just that our standards for replicability are
very high. We wanted to be comfortable that the data we actually
published appeared to be representative and robust--and we wanted to be
ready to replicate should Bellesiles ever release counts and cites for
any of the unpublished records he used.
We also thought it premature to recount any unpublished records that
did not cite and would not diclose his counts on. Why guess what he
included?--just stick with what we says he used (Providence and Jones).
Bellesiles would release any counts for particular counties in
records that actually support his low means, we would be more than happy
to study them as well.
As to Michael Bellesiles' claim to have used primarily the physical
records, rather than microfilm, that is a NEW claim he made only AFTER I
him in November that we were finding very different counts in the
records (and asked him to recount Providence, offering to lend him one
copies of those records).
Here is what Bellesiles said in an email to me from August 29:
>The statistics are in the appendix to my book [My note: only
counts, are >in the appendix]. But to answer your questions in a general
>let me note that I conducted the probate research before
>discovering the joys of statistical analysis on computers.
>The probate records are primarily on microfilm in the federal archives,
>though I also looked at several counties in their undigested form so as
>control for wills and any earlier transfer of guns. All of my note
>was on legal pads. I simply went through each probate record looking
>guns, recording each and every occurence of any type of gun or gun
>in any condition.
Note that he claimed to have looked at original records in only a few
counties") of the 32+ counties he studied in 1765-90. And the purpose
review was to supplement the count of guns from wills and earlier
guns (not to correct inventories in the microfilm). That the
essentially all on microfilm was repeated in Bellesiles' email to me on
13, when I first learned that his notes were damaged:
>Finally, all the original notes from this research are in boxes in my
>attic, where I moved them after they were damaged by a flood in Bowden
>Hall at Emory. I do not have the time right now to search for those
>and to lay out the water damaged pages for drying. However, I repeat,
>the material is easily available at any federal repository.
The federal depositories do not have this information, by the way. I
waiting for Bellesiles to get around to looking at his notes when I
late December that he had posted to this list the information that the
unrecoverable. Thus, only after I notified him in November that his
way off in Providence did Bellesiles claim that the records were
opposed to just wet) and that he relied primarily on original
than microfilm--original records being harder to check.
Further, it is not as if Bellesiles ever had a database to lose. He did
it all on
legal pads--over 11,000 ticks on legal pads.
Paul Finkelman also wrote:
> The data also requires careful reading. Is a "guon" of "gun"
> or a "gown" when it appears in a 17th century record? Simply reading
> written record from the 17th or 18th century is a skill that is not
> I think the very fact that historians recognize the ambiguity of some
> documentary records (and even thrive on it) is what makes it sometimes
> difficult for scholars in the two fields to communicate.
The idea that Alice Hanson Jones was unsophisticated enough to regularly
guns for gowns (sometimes spelled "gounes") is not plausible. These
transcriptions that are held in very high regard by serious quantitative
historians specializing in violence in early America, a fact confirmed
by such an
expert yesterday. Further, anyone who picks up these records will see
are usually listed with other weapons, or with other metal or household
items, while gowns are usually listed with clothes. If such errors
occurred, the percentage
differences would be slight. I saw the same sorts of language (guns,
in the microfilmed probate records from other years as Jones did in
is NO EVIDENCE that there are systematic errors in Jones' records
Note that Bellesiles offers no evidence for his speculations that there
transcription errors in Jones' work.
This is not a dispute that you have to take my word or Bellesiles' on.
up the Early Records of the Town of Providence (v. 6, 7, & 16) or Alice
Jones' American Colonial Wealth (1977/1978). In Providence, see all the
estates where they are supposed to be all male; see all the missing
they all are supposed to have wills; see all the guns that are not old
when most are supposed to be listed as old or broken. In Jones, skim
(69% guns) and you'll see that the pattern in 1774 is VERY different
from the one
Bellesiles reports for that region in 1765-90 (18%) using almost all the
Ultimately, this will be decided by scholars looking at the evidence.
Heather and I support all our claims with evidence, counts, and sources
Bellesiles has not yet cited a single source of records that supports
for probate records in the 17th or 18th century. Ultimately, those who
the evidence will decide this, not an internet discussion list guessing
arguments appear more plausible.
Professor of Law
Director of Faculty Research
Director, Demography of Diversity Project
Northwestern University School of Law
Chair-Elect, AALS Section on Social Science
Co-Founder, AALS section on Scholarship
Ph.D. Student (Sociology)
University of Chicago
I was trying to see how meaningful his data was, not show it was
completely inaccurate. When inaccuracies first appeared I began my
repeated efforts in private correspondence to have him supply data that
could be verified."
More information about the Conlawprof