Constructions, co-composition and merge

This article follows the line of research proposed in this volume by Mairal and Ruiz de Mendoza, Cortés Rodríguez, and Martín Arista, combining functional and constructional models of language. I discuss some interesting examples of break verbs with argument-adjuncts of motion which, being syntactically similar to Goldberg’s caused-motion construction, do not meet the definition for constructions devised by Goldberg. These structures are characterised by acquiring the semantics of motion without losing the semantics of change of state. Therefore, I have labelled them merge structures. Since Construction Grammar focuses mainly on constructions, I resort to Pustejovsky’s Generative Lexicon (henceforth GL) in search of a mechanism that allows me to explain this type of structure, because it is purposely designed to deal with creative uses of language and, more precisely, with polysemy. In a sense, it is not that different from Goldberg’s constructions, but it has the advantage that using the same system of lexical representation that is used for co-composition, it can also explain merge. The comparison of the two representations, co-composition and merge, indicates that the main differences can be explained by putting some of the semantic weight on the event and qualia structures. Although these representations capture the main differences existing between merge and co-composition, this argument only works at the level of interpretation or comprehension, a problem that was already pointed out by Goldberg (1995, 2006). Therefore, this article has also addressed the question of how to account for the production of those structures. Van Valin (in press) foresees the possibility of reconciling projectionist accounts like Role and Reference Grammar (RRG) and constructionist approaches like Pustejovsky’s GL and argues that the interpretive role of co-composition is helpful in the linking from syntax to semantics in RRG. Thus, when presented with a certain syntactic structure, the choice between a merge or a co-compositional reading will be determined by their lexical representations.