140 andrew w. pitts
these are ranked from top to bottom (starting with individual) and from
bottom to top (starting with group) according to the highest levels of lin-
guistic impact, based on current research, so that the graph can feature
the greatest spikes in social variation along the dimension of co-textual
(statistically weighed) spikes. Since we have some variation within the
involved-information production scores within the Pauline corpus, i will
just focus on analysis on the three dimensional shift in addressee type.
these three components also have far more data accrued in sociolinguis-
tic studies so that the correlations made here will be most convincing. in
order to gauge the correlation between co-textual and contextual varia-
tion according to these margins, we merely extract involved-information
scores and plot statistical co-textual variation against this graph of social
variation to see whether the levels of variation are consistent with what
we tend to find in sociolinguistic studies of single author style-shift within
a corpus involving major shifts in addressee type. the vertical co-textual
column allows for the insertion of statistical frequencies in terms of per-
centages for co-textual variation, such as vocabulary.
We turn now to consider how a register design interpretation of linguis-
tic variation renders the data usually emphasized by the pseudonymity
view, especially semantic clustering and co-textual variation. a significant
weakness of the pseudonymity interpretation is that it only considers the
co-textual axis of variation. a register design (or almost any contemporary
sociolinguistic) model emphasizes the importance of correlating this level
of variation with the contextual axis.
A Register Design Interpretation of Semantic Clustering
While many co-textual factors unify the various register profiles within
the Pauline letter-writing register, we find many other features that have
seemingly random patterns of divergence. the article, for example, does
not seem to have a particular pattern within any one register profile.70
there is also a great deal of divergence in grammatical case usage that
cannot be restricted to clustering in any of the traditional sub-corpora of
the Pauline corpus. the same is true of gender and number distribution
as well as the ratio of nouns to pronouns. nevertheless, several seman-
tic clusters do emerge and this clustering phenomenon has been subject
to all kinds of interpretations from literary dependence and/or fragmen-
tary hypotheses (e.g., the colossians-ephesians relation) to pseudonym-
70 cf. Kenny, Stylometric Study, 85.