This article examines instances of the pu relative construction in Modern Greek in which the semantic role of the head is underspecified by the syntax. Such cases include sentences whose nominal head corresponds to some complement of the relative clause predicate and sentences in which the head does not have any sort of syntactic relationship with the relative. The latter, which are characteristic of oral, informal discourse, have been completely ignored in the previous literature, which has defined relatives on the basis of exclusively structural criteria. It is argued that a unified account of the pu-construction (including gapped and gapless relatives) can be achieved if we analyze it as a conventional instruction for a particular kind of conceptual integration. Semantic and pragmatic factors influencing successful construal (one which leads to the construction of a unique blend) are systematically examined. The lack of a clear cut-off point in acceptability for such utterances tallies with the conclusion reached here, namely that the constraints governing such uses are constraints on interpretability.
In this paper we address lexical polysemy in a constructional perspective, arguing that each of the conversational meanings we identify for Modern Greek ela (2nd person singular imperative of the verb erxome ‘come’) is appropriately modeled as a conceptual gestalt of formal (including prosodic) and semantic-pragmatic properties. In turn-initial position, ela is used to challenge a preceding utterance; we show that the variations in the kind of challenge expressed are systematically tied to the word that follows ela, the speech act force and the sentence type of the preceding utterance, and finally prosodic and textual cues. To the extent that these varieties of conversational challenge are conditioned by particular contextual features, we treat them as a family of related constructions whose common features can be captured in the form of a generalized ela construction abstracted from the different sub-patterns. Our analysis thus demonstrates the appropriateness of a constructional framework for dealing with the different kinds of parameters involved in dialogic meaning and strongly suggests that at least some of the variation inherent in discourse is amenable to a grammatical description, so that sentence-level and supra-clause patterns can be analyzed in a uniform way.