Patten (2010) and the constructionalization of the English it-cleft
In Patten (2010), I identify some problems with Ball's (1994) diachronic account. I suggest that while the it-impersonals certainly show a superficial similarity to it-clefts with adverb phrase and prepositional phrase foci, they have little in common (structurally or functionally) with the existing NP-focus it-cleft. This raises questions over why these two configurations would merge together. Ball's second merger, which results in the development of NP-focus it-clefts with new information in the cleft clause, evinces similar problems. Ball (1994: 621) acknowledges that her account, involving the merging of multiple constructions, is somewhat ad hoc; she concedes, "It is unlikely that we shall ever be able to pinpoint the cause of the appearance of the IP NP-focus it-cleft". Furthermore, none of Ball's source constructions can account for the peculiar function of the IP it-cleft. In IP it-clefts, the cleft clause expresses hearer-new, yet presupposed, information. For instance, in (24) above, the audience may not be aware that someone said 'there are no facts, only interpretations', but there is nevertheless the expectation that they will accept this information as a generally known fact (see Prince 1978). However, while the it-impersonals present new information which is not presupposed, reverse pseudoclefts contain presupposed information which is used "for re-evoking information, rather than communicating entire new propositions" (Oberlander & Delin 1996: 200).
I go on to re-examine the development of the English it-cleft from a constructional perspective. Supplementing Ball's findings with data from the Penn Corpora of Historical English, I find evidence of a gradual progression, in which it-clefts occur with an increasingly wide range of focal categories, which are less well suited to performing the requisite referring function. Originally focusing NPs referring to the most discrete of entities, the it-cleft expands to sanction prepositional phrase foci (which relate to nominal concepts), before occurring with adverb phrase and clausal foci, before finally accommodating phrasal categories which resist a referential reading (such as adjectival foci), shown in (26).
(26) It's not sick that he was but tired (E. Kiss 1998: 262)
Likewise, the development of the IP it-cleft seems to occur in incremental stages. Over time, the it-cleft construction changes from expressing only given information in the cleft clause, to expressing shared (yet non-salient) information, to accommodating hearer-new but nevertheless factual information, before finally allowing uses where the speaker incorporates their subjective opinion into the presupposition.
As a result, I claim that the non-NP-focus it-clefts originated by extension from the existing it-cleft tokens and were not directly influenced by the OE impersonal construction. This is consistent with a usage-based, constructional model of language change (see 2.2). Within this framework, the development of the English it-cleft can be interpreted as an example of schematization (or grammatical constructionalization). The change begins as speakers extend the existing it-cleft schema to form novel instances. As new and old instances coexist, speakers generalize over them, forming abstractions which stipulate only those characteristics that are shared by all of its members/instances. For example, as speakers abstract over instances with both NP and non-NP foci, the overarching schema loses a syntactic constraint and no longer specifies for a postcopular NP. However, the it-cleft retains a semantic condition such that the postcopular XP must refer. It is this property that governs the course and rate of the development, since some syntactic categories (and members of these categories) are better suited than others at performing a referring function. Likewise, the accommodation of hearer-new information into this presuppositional construction becomes conventionalized as speakers abstract over these instances to form a new sub-construction: the IP it-cleft (see also Lambrecht 1994; Lehmann 2008). This, in turn, has implications for the overarching it-cleft category, which evolves into a more productive, and therefore a higher-order, construction. The gradual trajectory of this change is governed by the degree to which new instances override the more general pattern of correspondence between discourse-old and presupposed information.
On this account, the development of the it-cleft construction is both gradual (occurring in incremental stages) and directional (proceeding upwards throughout the hierarchical system). This interpretation fits with other examples of constructional change that have been explored in relation to grammaticalization phenomena. Constructions show a tendency to become more schematic (or open), more productive (sanctioning more instances) and more conventional (developing construction-specific correspondences) over time (see Traugott 2007; Trousdale 2008, 2012, this book). From this perspective, the specificational ¿t-cleft has undergone a process of context-expansion: both structural 'host-class' expans¿on, whereby the construction allows a wider range of components to enter into it, and semant¿c-pragmat¿c context expans¿on, through which the construction develops new pragmatic functions (see Himmelmann 2004). The focus here is on how these changes progress, via analogy (or extension), rather than on their initial actuation.
I suggest that this approach better suits the diachronic data, in that it accounts for the continuing (and perhaps ongoing) expansion of the ¿t-cleft construction beyond LME. It also anticipates that any changes impact on the construction as a whole. For instance, the changes to the postcopular XP not only affect the new AdvP/PP-focus ¿t-clefts, they also influence ¿t-clefts with NP-foci: my data shows that the NP-focus ¿t-cleft occurs with more abstract foci over time, and begins to occur with time and place adverbials in the focus position at the same time as these are found in the AdvP/PP-focus ¿t-cleft. This suggests that the change does not apply wholesale to discrete syntactic categories; instead, it is governed by existing semantic overlap between these categories and by variation within those categories. The constructional framework encourages an approach which relies less on discrete syntactic categories; since on this model, the construction is the primitive, categories derived from their distribution across constructions are expected to exhibit variation (Croft 2001, 2007b).
This diachronic account is also compatible with the synchronic, extraposition-from-NP analysis of ¿t-clefts that I adopt (see 3.2). On this analysis, the cleft clause is treated as a restrictive relative modifying the initial ¿t. Consequently, the syntactic category of the focal element does not influence this analysis in any way. For instance, the PP-focus ¿t-cleft in (22) above (repeated here as (27)) is likewise analysed as containing a determinative (restrictively modified) pronoun, forming a discontinuous definite NP: ¿t (the t¿me) that he starts play¿ng. On this extraposition account then, ¿t-clefts with non-NP foci do not form a separate construction and so do not require a separate historical source from NP-focus ¿t-clefts (contra Ball 1994).
(27) [It ]'s at 2.30 [that he starts play¿ng]i [PP-focus ¿t-cleft]
While Ball does not really explore the possibility that tokens with non-nominal foci could have resulted solely from the gradual expansion of the ¿t-cleft construction, she does discuss (and reject) the suggestion that the IP ¿t-cleft evolved "by a progressive weakening of discourse requirements" (Ball 1994: 615). Ball (1994: 615) observes that while there are some early examples of an "intermediate stage", in which ¿t-clefts express shared (yet non-salient) information in sentence-final position, in her data "they occur in translations of Latin pseudoclefts...[and so] are not by themselves reliable evidence for pragmatic change". Ball's (1994) objection is a fair one. While in my data, not all of the shared (yet non-salient) examples are from Latin translations, it is true that in Middle English, the z'f-cleft is still a relatively low frequency construction and the IP z'f-cleft data set is very small. Nevertheless, the fact that, post-LME, IP z'f-clefts begin to occur with greater frequency (suggesting that the new discourse-function has become conventionalized) and we find evidence of a further manipulation of presuppositional meaning (as speaker opinion is accommodated into this position), certainly seems to me to be indicative of grammatical constructionalization.
Furthermore, it interests me that while Ball objects to the use of Latin translations as evidence for analysing the zf-cleft's development as a case of gradual expansion, her account of the z'f-cleft's origin is entirely dependent upon a rare configuration, found only in translations from Latin originals (see 4.1). The way that Ball and I interpret, and evaluate, the historical data is therefore very much influenced by our theoretical assumptions; our ideas about which evidence counts as relevant and which can be overlooked are more or less governed by how well it supports our expectations about how the language is structured and how languages change.
-  Although this section is based exclusively on Patten (2010), I return to the diachronic development of if-clefts in Patten (2012). In extending the dataset (beyond Late Middle and Early Modern English), I have deepened my analysis. Consequently, some of the details differ.