5.6 Development 157Keeping the Old English situation in mind, in the sections that follow I will con-
sider both syntactic and semantic aspects of the development of epistemic paren-
theticals, arguing that their development provides a crucial test case for the matrix
clause hypothesis (see Section 1.4.1.2 ) and can be considered a case of grammati-
calization, albeit not a “textbook case” (see Thompson and Mulac 1991 : 324– 325).
5.6.2 Syntactic Development: From Main Clause to Parenthetical
According to the matrix clause hypothesis (see Section 1.4.1.2) , epistemic par-
entheticals ( I guess , I think , I suppose ) are postulated to arise as matrix clauses
with dependent that - complement clauses. Deletion of the complementizer that
creates an indeterminate (or “bridging”) structure where the original matrix
clause can be reanalyzed as parenthetical and moved to medial or fi nal posi-
tion, as shown in the sequence: I think that the world is fl at > I think Ø the
world is fl at > The world is fl at, I think. According to Thompson and Mulac
( 1991 ), evidence for this scenario is quantitative. Verbs which give rise to epis-
temic parentheticals are among the most frequent that- clause complement- tak-
ing verbs, and more importantly, they occur overwhelmingly with fi rst- person
subjects and without that. Thompson and Mulac ( 1991 : 324) admit, however,
that the process they describe “is largely attested only synchronically” (324).
Can the situation Thompson and Mulac fi nd in Present- day English be
extrapolated back to Middle English to account for the rise of epistemic par-
entheticals in Middle English? In her study of the rise of methinks , Wischer
( 2000 : 361) suggests that it can. She points out that methinks is complemented
by a that - clause and that that is often deleted; as a consequence, the former com-
plement clause can be reanalyzed as a main clause and the impersonal methinks ,
formerly the main clause, as a subordinated disjunct.^26 But Wischer provides
no quantitative evidence, thus leaving open the question of whether the Middle
English data show the quantitative correlation Thompson and Mulac found in
Present- day English between that - less forms and epistemic parentheticals.^27
We fi nd all three PDE structures identifi ed by Thompson and Mulac ( 1991 )
in Middle English, including both medial and fi nal parentheticals:
(24) a. I woot right wel that swich was my biheste. (1387– 1400 Chaucer, CT
D.WB 1059)
‘I know quite well that such was my promise’
26 Wischer admits ( 2000 : 366n.) that further study of “the general relationship between the dele-
tion of the subordinator that and a probable reinterpretation of main and subordinated clauses
in English” is necessary. See Palander- Collin ’s quantitative studies ( 1996 , 1997 , 1999 ).
27 We must keep in mind that Thompson and Mulac ’s results are based on conversational, oral
English, while, obviously, the Chaucerian data are based on written texts, although with a fairly
high proportion of represented speech.