Esitelmä 1. virolais--suomalaisessa kognitiivisen kielitieteen konferenssissa Taageperan linnassa 13.12.2006
Some practitioners of Construction Grammar, most notably Kay (2002), make a fundamental distinction between on one hand constructions, which are seen as parts of the grammar, and on the other hand patterns of coining, which are used to extend the grammar. However, it is not clear that such a sharp distinction is justifiable, and in fact Fillmore et al. (1988) seem to imply that this distinction is not a sharp, qualitative one. In an attempt to explain similar phenomena, Leino and Östman (2005) have proposed metaconstructions as a way to specify tendencies in grammatical organisation.
I am looking at this issue from the perspective of place names. A study of the co-location patterns in a large computerised toponym corpus reveals numerous pairs such as Mustalampi `Black Lake' - Valkealampi `White Lake' or Ahvenlampi `Perch Lake' - Haukilampi `Pike Lake', where the names of nearby lakes are clearly related to, and influenced by, each other. It is evident from the amount of such pairs that the same innovation is made time and again; these names cannot be considered results of singular events, except in the sense that any name-giving is a singular event. This indicates quite clearly that productivity is not an on/oc matter but rather a sliding scale, like so many other distinctions in language.My goal is to outline an explanation for both the phenomena involved in contrastive naming and those that have led to the theory of metaconstructions. To do so, it is necessary to weaken the distinction between generalised constructions and specific constructs: from the onomastic analysis it is clear that existing constructs occasionally act as prototypes in creating new ones. This must, of course, be restrained in order to keep the explanatory power of the grammar, but such restraints can be devised using criteria like entrenchment and salience (see e.g. Giora 1999), much in the same way as Goldberg (1995) finds correlation between the type frequency and productivity of an inheritance link.
It is also necessary to make some adjustments to the way existing elements are integrated with each other. Unification, at least in the strictly formal sense, is hard to reconcile with a prototype-based approach. However, combining constructions to create new constructs can be described without the use of formal unification: instead, the process can be explained in terms of conceptual integration (e.g. Coulson and Oakley 2000).