20 Pragmatic Markers: Synchronic and Diachronic
matrix clause with that- less complement, at the time at which the parenthet-
ical can be seen to develop argues against the Thompson and Mulac proposal,
which is predicated on the frequency of the source construction for the rise of
the parenthetical (see Brinton 2006 : 249– 250).
Moreover, the history of that - deletion is by no means a straightforward case of
loss. A number of syntactic, discursive, and stylistic factors are known to favor
the retention of that , including complexity of the matrix clause, elements inter-
ceding between the matrix verb and that , full NP subjects in the complement,
and so on ( Rissanen 1991 ; Finegan and Biber 1995 ; Biber et al. 1999 : 680–
683). There are also signifi cant differences in that- deletion among different
verbs. Rissanen ( 1991 ) shows that the highest rate of zero- complements was
in the second half of the sixteenth and part of the early seventeenth century,
but this was reversed in the “norm- loving” eighteenth century. In fact, Finegan
and Biber ( 1995 ) fi nd a consistent rise in the use of that from 1650 to 1990.
Comparing a number of corpus studies based on different genres and corpora –
studies which prove to be remarkably consistent – Torres Cacoullos and Walker
( 2009 : 5) confi rm these results, i.e., that the frequency of zero- complements
reached its peak in the middle of the seventeenth century only to fall quite dra-
matically thereafter. The commonplace that that - deletion has increased over
time as English has become more “colloquial” ( Mair 2006 ) is thus not found
to be the case, though oral genres favor it ( Rissanen 1991 ). Prescriptive tenets
requiring the use of that may have disrupted the natural trend.
In recent years, a series of studies have argued – on synchronic grounds – that
the presence or absence of that does not distinguish between main- clause and
comment- clause uses of initial I think / I guess forms. Kearns ( 2007 ) proposes
that the loss of that is not a sign of syntactic reanalysis but correlates with a
(pragmatic) change in “informational prominence,” where the embedded clause
becomes more prominent than the matrix (e.g., the matrix can be omitted with
altering the propositional content of the main assertion, the matrix can be para-
phrased with an epistemic adverb , the matrix corresponds to clause- medial or
fi nal forms, and tags are determined by the embedded clause).^22 Whether or not
that is present does not correlate with the different uses of I believe or I think ,
according to Dehé ( 2010 ). What distinguishes the use is prosody. Main- clause
(assertive, propositional) uses are characterized by accent on the pronoun; the
sequence may be a separate intonation domain or join the following domain.
Comment- clause uses, whether prosodically integrated or not, have accent on
the verb; that may join the pronoun– verb sequence (suggesting a high degree
of routinization with that not acting as a complementizer) or the following
domain. Discourse- marker uses appear to be those in which I believe/ think are
22 Kearns ( 2007 ) notes that the epistemic forms are compatible with third- person subjects, com-
plex verb phrases, adverbials, and past tense (see also Van Bogaert 2010 ).