Skip to Content.
Sympa Menu

corpus-paul - Stylostatistics

corpus-paul AT lists.ibiblio.org

Subject: Corpus-Paul

List archive

Chronological Thread  
  • From: "Mark House" <mhouse AT fuller.edu>
  • To: "Corpus-paul" <corpus-paul AT franklin.oit.unc.edu>
  • Subject: Stylostatistics
  • Date: Mon, 10 May 1999 07:34:59 -0700


Greetings to all.

I'm new to the list, and have enjoyed following the threads so far. I've
especially appreciated the discussions of stylostatistical research, which
is the topic of my current research. Many thanks to Sheila McGinn for
sharing her work on Neumann.

Another scholar doing recent stylostatistical work on the Pauline epistles
is David Mealand. His earlier work seems to corroborate Neumann's
skepticism regarding Morton's rejection of Philippians and the Thessalonian
epistles as authentically Pauline. (He published another study in 1995
using multivariate analysis, which I'm still working on.) Here's what I put
together in response to some of his earlier work. It too points out the
danger of relying on single variables to discern authorship.

Pardon the ususual transliterations of Greek words. I didn't bother to put
them into B-Greek form, but I think they're obvious.

=========================

Review of David L. Mealand, "Positional Stylometry Rreassessed: Testing a
Seven Epistle Theory of Pauline Authorship," XXXV (1989), 266-286.

Mealand1 begins by noting recent ways in which older measurements of style
have been supplemented. Old word-count based assessments have been enhanced
with calculations of "the frequencies of certain particles, sentence
lengths, use of the subjunctive, use of specific tenses and the like."2 Yet
despite the use of increasingly refined methods, scholars have arrived at
vastly differing conclusions with respect to the Pauline authorship of the
disputed epistles of the NT. The cautious approach of A. Kenny led him to
regard twelve of the thirteen Pauline epistles as the work of "a single,
unusually versatile author."3 A. Q. Morton, on the other hand, regarded
only four of the epistles as genuine works of Paul.4

Neither of these conclusions matches "the mainstream of literary critical
study of the Pauline epistles" which regards the Pastorals as later
productions, with some disputing Ephesians and Colossians as well.
Beginning with the seven epistles generally accepted as Pauline by literary
critical scholars,5 Mealand proposes to run new tests comparing these seven
with Morton's four and a broader list of ten.6 Using these new tests, he
hopes to provide not only overall results, but intermediate figures on each
of the individual books in question.

The method Mealand uses to run these tests is known as positional
stylometry, which is a method based on "the relative frequency of particular
words in particular positions in the 'sentence'."7 Although he concedes
"the controversial character of stylometry in general," and positional
stylometry in particular, he challenges the method's critics to give a
"clear-headed assessment of just what it is about the method" that they
contest.8 In light of the scrutiny of the critics, he vows to pay careful
attention both to "the logic of the argument and to the pool of assumptions
used."9

In these tests, Mealand measures the common words kai', de', ga'r and ei> as
they occur in the first, second, second and first positions of the sentence
respectively.10 Beginning with similar studies done by Morton, Mealand
revises the figures where needed,11 and performs fresh c2 tests12 on the
data to arrive at significance levels of diversity within groupings of
epistles, ranging from no significant diversity to extremely significant
diversity for a given test. The higher levels of diversity would then be
used, when compiled with other data, to assess the likelihood that a
particular writing is "heterogeneous" -- the result of a different author.

Having laid out his method, Mealand goes on to discuss the results of the
tests for each of the four categories of words. Using kai' as an indicator,
he concludes that there is "no good reason on this test to regard the four
epistles as heterogeneous," since none of the variations in the use of kai'
between the four are of high levels of significance.13 When the ten
epistles are tested together, however, highly significant diversities
appear, leading to the judgment that "the sample of ten epistles is
diverse."14 When the grouping of seven epistles is tested, the results are
"more interesting and less clear-cut."15 Although the differences between
all seven are considered marginal, "if the group is diverse...the source of
the diversity is to be found in Galatians or 2 Thessalonians."16 Mealand
suspects that 2 Thessalonians is the culprit, due to its brevity.

Mealand's conclusions agree with Morton's in affirming the coherence of the
four epistles and in seeing a clear pattern of diversty among the ten.
However, his tests conflict with Morton's earlier results in that Mealand's
results show that with an intermediately sized group (seven or six
epistles) "it is by no means clear that that group is diverse."17 Thus
Mealand's results agree more closely with conventional literary critical
studies.

Mealand next discusses test results with de' as the second word in the
sentence. These are somewhat at odds with his kai' tests in that Galatians
has "a markedly high incidence" of the former when put into a group of just
four epistles. Since Galatians is among the least disputed books according
to traditional literary criticism, Mealand accounts for this high number on
the basis of "the acutely argumentative character of Paul's relations with
the Galtians." Interestingly, when set in a group of seven epistles,
differences between the letters in their use of postpositive de' are only
marginally significant, suggesting to Mealand that "the seven form a
coherent group."18 A further test in which Galatians (which has an
unusually high incidence of postpositive de') was omitted reduced the level
of diversity among the seven to an insignificant level. Again, Mealand
appeals to the "sharply polemical character" of the Galatian situation to
account for its diversity.19 Tests comparing ten epistles once again showed
high levels of diversity, suggesting a lack of homogeneity of authorship.

The results of tests done on ga'r as the second word in the sentence were
"markedly different from the results of the first two sets of tests."20
Testing all three groups of four, seven and ten, Mealand finds in every case
significant levels of diversity. This result, he comments, "should be as
surprising for the upholder of a four epistle theory as it is to me with my
declared interest in testing a seven epistle theory."21 To explain this
discrepancy, Mealand conducts further tests on the controls used in the
experiements. Using the works of Isocrates which Morton had tested earlier,
Mealand finds a high degree of diversity in the use of secondary ga'r there
as well. He suggests that this diversity may be due to changes in literary
genre as well as "subject matter, situation, mood, and the capacity for
rational thought of author and readers."22 Testing literary genre in
Isocrates, he discovers that the author's political works, when separated
from his epideictic writings, show a marked decrease in diversity in their
use of ga'r. Similar tests with respect to "subject, situation, mood" are
called for with respect to the Pauline epistles if these diecrepancies are
to be resolved.23 For the time being, ga'r, it would seem, is an unreliable
indicator of authorial diversity.

Mealand's final test concerned the use of ei> in the primary position. Here
again, the results were significantly different from those of the others
three tests. Whereas with ga'r significant levels of variation were found
with all three sets of epistles, just the opposite result occurred with ei>.
None of the three groups showed any significant variation. Taken by
themselves, the ei> tests would suggest that all ten epistles were
homogeneous. Since this result conflicts with the other tests, Mealand
suggests that perhaps this word is a"poor discriminator"24 of authorial
diversity. A further test bolsters this conclusion. Mealand randomly
chooses an oration of Isocrates and compares the frequency of its use of ei>
with that of Romans, finding again no significant diversity.

In the "Conclusions" section of his article, Mealand deduces that using
positional tests of style "a ten epistle theory comes out badly" with
respect to homogeneous Pauline authorship, while a seven epistle scheme is
nearly as strongly unified in style as Morton's four epistle model.25 The
relatively small degree of diversity that does exist between the seven is
best explained by "differences in situation, genre, mood or whatever."26
While he admits that "one of the tests fails to discriminate sufficiently"
and that another "shows too much variation," he judges that his evidence
"tells against a ten epistle theory" of Pauline authorship and leaves a
seven epistle theory as probable as a four epistle model.27

Looking at Mealand's results as a whole, we might be less inclined to see
the significance of his tests. While they do indeed seem to cast doubt on
the validity of Morton's conclusions, they are hard pressed to yield
confidence in positional stylometric methods in general. Several factors
seem to support such skepticism:

1) Only the kai' tests consistently support his conclusions by showing
great diversity with ten epistles, insignificant diverrsity with four, and
only negligible diversity with seven. None of the other three tests in
themselves lead consistently to this conclusion.

2) Two of the four items tested (ga'r and ei>) showed results divergent
enough from the other tests to be considered poor indicators of authorial
diversity. This effectively reduces the weight of Mealand's evidence by one
half. It also at least raises the quesion of the validity of any of these
common words as good indicators of an author's distinctive style. Could the
positive result from the kai' test be due to mere coincidence?

3) With the de' tests, Mealand's results called into question the Pauline
authorship of Galatians, which has been unquestioned not only by the bulk of
critical literary scholarship but even by one as skeptical as Morton. This
forced Mealand to appeal to a host of other factors that might account for
the stylistic diversity found in Galatians. If the method used casts doubt
on a mainstay Pauline epistle such as Galatians, does this not raise
questions about the value of the procedure as a whole?28

Mealand's later work moves in the direction of seeking a profile of
authorial style by using a multitude of stylistic indicators.29 We can only
surmise that the inconsistency of results using just a few indicators has
caused him to continue his search for more solid methods. But the results
of these earlier tests raise again the question of some of stylometry's
harshest critics: Are stylometric arguments strands which together form a
strong cord binding a particular work with a known author, or do they rather
form a chain which is only as strong as its weakest link? Even using the
rope metaphor, Mealand, in these tests, gives us precious few strands of
evidence upon which to hang our judgments regarding Pauline authorship of
disputed epistles.

Notes:

1 Mealand, David L. "Positional Stylometry Rreassessed: Testing a Seven
Epistle Theory of Pauline Authorship," NTS XXXV (1989), 266-286.
2 Ibid., 266.
3 A. Kenny, A. A Stylometric Study of the New Testament (Oxford: Carendon,
1986), p. 100. Kenny regarded only Titus as incompatible with Pauline
style.
4 A. Q. Morton, Literary Detection: How to Prove Authorship and Fraud in
Literature and Documents (New York: Scribner, 1978), 168, 183. Morton
accepted Romans, the two Corinthian epistles and Galatians as Pauline.
5 Rom., 1 and 2 Cor., Gal., Philip., 1 and 2 Thes. Philemon is omited due
to its brevity.
6 Mealand, 167. The broader list includes Eph., Col., and 2 Tim. (the
Pastoral most widely held to contain Pauline elements).
7 Ibid., 268. A "sentence" (referred to by Mealand as a c-sentence) is a
group of words ending not only in a full stop or question mark, but also in
a colon.
8 Ibid.
9 Ibid., 269.
10 These are designated k1, d2, g2 and e1 respectively.
11 Figures may differ slightly depending on which electronic Greek text is
used for the computation. Morton, Friberg and Gramcord are three commonly
used examples. Morton also notes errors in previously used figures which
have crept into subsequent analyses.
12 The "chi-squared" (c2) test is a standard statistical method for
assessing the significance of a difference between observed counts and
expected counts of a given variable. Mealand bases his counts on Morton's
Greek text, but corroborates his findings using Friberg.
13 Mealand, 277.
14 For this test, 1 and 2 Thessalonnians had to be combined to yield numbers
of sufficient quantity to be tested.
15 Ibid., 278.
16 Ibid. In Gal., kai' occurs in the first position relatively
infrequently. In 2 Thes., just the opposite is the case. This result is
interesting in light of the coherence between Gal. and the other three of
the four epistles.
17 Ibid., 279.
18 Ibid., 280.
19 Ibid.
20 Ibid., 281.
21 Ibid.
22 Ibid., 282.
23 Ibid.
24 Ibid., 284.
25 Ibid., 285.
26 Ibid.
27 Ibid., 286.
28 This is particularly the case when one considers that Galatians is one of
the four epistles used to establish the very contours of Pauline authorial
style.
29 See epecially "The Extent of the Pauline Corpus: A Multivariate
Approach," JSNT, No. 59 (1995), 61.

Mark House
Ph.D Candidate
Fuller Theological Seminary
Pasadena, California




  • Stylostatistics, Mark House, 05/10/1999

Archive powered by MHonArc 2.6.24.

Top of Page