Cross-domain variation in the X itself as a grammatical construction

The X itself is a nominal construction that has not received much attention within cognitive linguistics despite it having a quite interesting function, as it serves to select a core part in a partonomy and thus specify lexical relations within a text. Apart from being mentioned in passing in Croft & Cruse (2004), one of the few treatments of this construction in cognitive linguistics is Jensen (2014) who builds on the comments in Croft & Cruse (2004) and proposes a hypothesis pertaining to the cognitive and discursive function of the construction. However, that hypothesis does not take into account an important aspect of the reality of language — namely, variation. This article investigates, within the framework of usage-based construction grammar, the X itself in the Open American National Corpus (OANC) to see whether the construction displays variation across the nine domains that the data in OANC are divided into. Applying quantitative techniques, including lexical diversity measures and multidimensional scaling, this article explores aspects of the discursive behavior of the X itself across these domains and addresses the extent to which the construction interacts with the registers associated with the domains. Focusing on use-based varieties (McArthur 1992, see also Quirk 1989 and Halliday et al. 1964: 77), the present article argues that the X itself is not a constructional monolith, but that it is characterized by register-sensitive functional variation and that its core selection function very likely serves a information-structural discourse-pragmatic purpose.