BEST PRACTICES 367
McMenamin evaluated as “odd” such spellings as [abit, a lot,
anytime]. But when Goutsos used a typical linguistic
methodology of checking for frequency in a corpus, in this case
the ten-million word corpus of American English, the Bank of
English Database, he found such spellings sufficient to
comment: “this would imply that careful research must precede
any prescriptive judgment.”^87 Indeed.
Certainly, in Professor McMenamin’s defense, his later book
includes a chapter in which he does consider statistics that could
be used in a forensic linguistics analysis.^88 Further, he does
write about a corpus he is developing.^89 But there is still a real
gap between the theory put forth in the book and the method and
conclusions put forth in Professor McMenamin’s actual analyses
and reports, as shown by Nunberg’s peer review.^90
Nunberg prepared an affidavit in which he stated:
I believe I have a responsibility as a linguist to point out
the deficiencies of Dr. McMenamin’s work, which
misrepresents the methods of the discipline of
linguistics....
- Professor McMenamin’s methods are not based on
well-established theoretical principles nor are they
consistent with rigorous practice in the statistical analysis
of written texts. McMenamin has performed no statistical
research that would give any scientific grounding to his
conclusions. I would not classify McMenamin’s work as
bad science; rather, it is not science at all. - Professor McMenamin’s choice of the features used
in document comparison is arbitrary and subjective, and
unmotivated by any empirical research; another set of
features could well have been chosen that would have
given very different results. His method could not pass
the test of independent replicability.
(^87) Id. at 105–06.
(^88) MCMENAMIN, ADVANCES, supra note 64.
(^89) Id.
(^90) Statement of Geoffrey Nunberg, In re Marriage of Hargett, No. SDR-
0017114 (Cal. Super. Ct. 2005) (on file with author).