The grammaticalization of I think has received considerable attention in recent years (Thompson & Mulac, 1991; Kaltenböck, 2011; Van Bogaert, 2011). However, far less attention has been paid to the related epistemic marker I’m sure, despite the fact that this construction has been shown to exhibit similar behavior (Kaatari, forthcoming). The present study aims to investigate the degree to which I’m sure is on the same grammaticalization trajectory as I think, as explained below.

Following Traugott & Heine (1991), we view grammaticalization as both a diachronic and a synchronic phenomenon to be studied “at a synchronically segmented moment in time” (Traugott & Heine, 1991:1). In addition to a propensity for that-omission in clause-initial position, one of the main arguments put forth to support the claim that I think is grammaticalized is that it has developed an ability to occur in clause-medial (1) and clause-final position (2), that is outside its canonical clause-initial position (3) (Thompson & Mulac, 1991). Contrary to Hooper’s (1975) claim, a recent empirical study on I’m sure has indicated increased flexibility in this respect, as exemplified below (Kaatari, forthcoming).

(1)  He is, I think/I’m sure, an interesting person.

(2)  He is an interesting person, I think/I’m sure.

(3)  I think/I’m sure (that) he is an interesting person.

Nonetheless, the question remains whether the development of the two constructions can be accounted for in the same way, despite the fact that these constructions have different frequency entrenchment and that the predicates belong to two different word classes.

The aim of the present study is to investigate whether I’m sure follows the same grammaticalization trajectory as I think. The research questions are as follows (see Lehmann, 1985:303, for a discussion of the methodological parameters of grammaticalization used):

• What is the frequency distribution across the clausal positions (syntagmatic variability)?
• To what extent is the complementizer that omitted (paradigmatic variability)?
• Are there any differences across time such that the development of I’m sure could be considered to mirror that of I think?

The study uses comparable subsets from the spoken component of the BNC (Burnard, 2007; Lee, 2001) and the newly compiled Spoken BNC2014 (Love et al., 2017). The results show that I think and I’m sure exhibit remarkable similarity, especially in the most recent data, not only in terms of their proportional distribution across clausal position, but also in terms of their propensity for that-omission. Even though the time span covered is relatively short, a clear increase of that-omission can be noted for I’m sure, for which the frequency increased from 93 to 98 percent, thus mirroring the frequencies for I think (99 percent) very closely. In order to reconcile the fact that I think and I’m sure thus exhibit similar behavior, despite differences in frequency entrenchment, we argue that both constructions are part of the same constructional grammaticalization schema in which the frequency of I think seems to reinforce the grammaticalization of both I think and I’m sure.