Syntactic constructions are known to make a genuine contribution to sentence meaning, and this construction meaning must be integrated with lexical meanings to yield a sentence’s meaning. We focus on the variability of this lexeme-construction integration and especially in cases where conceptual enrichment, based on background or contextual knowledge, is necessary in order to achieve, for instance, argument linking. We will undertake several empirical case studies in order to investigate the potential and the limits of such enrichment processes. Our final goal is the development of a model of syntax-semantics mapping that is more flexible than the models assumed so far and includes an explication of its relation to non-linguistic meaning – the situation model – both in interpretation and production of linguistic expressions.

We shall develop a formal model of a linking mechanism that uses three levels: syntax, semantics and situation. The need of a distinct level — a situation model — will be argued for on several grounds, both theoretical and experimental. Theoretical arguments come from the impossibility of defining linking properties solely in terms of semantic representations (which are abstract) or solely in terms of the situation model. We hope that experimental arguments will also be provided which will show that the actual linking that is chosen by interlocutors depends on both levels not just one. From a theoretical point of view, the experimental arguments presuppose that the levels can be successfully differentiated. In particular, we shall investigate the following questions:

  1. What is the relationship of the three level model to standard two level approaches?
  2. Is it possible in principle to reformulate conditions on situations as conditions on representations or vice versa? If so, are these reformulations natural or artificial?
  3. Can the need for three levels successfully be shown not only for representational purposes but also for linking purposes?
  4. How much interaction is there between the three levels?

Furthermore, we shall look at the role of frames in the semantics. For example, we shall investigate whether frames encode linguistic knowledge or whether they do, in fact, code properties of the situation model. Further questions to be discussed are the role of background knowledge (physical, cultural or otherwise) in the linking process. Finally, from a theoretical perspective we need a clear definition of the term “construction” so as to be able to engage in a fruitful discussion with other theoretical models. It is necessary to ask, for example, how linking is done in terms of constructions: For it appears from the literature that constructions themselves are items and do not perform the linking themselves. Hence an additional mechanism of linking is needed. In this connection we need to see whether enrichment is part of the general linking algorithm or whether it is construction specific. If the first is the case enrichment must be thought of as a general, automatic cognitive mechanism. If the second is the case enrichment is, at least in part, not automatic.

We will carry out four empirical case studies concerned with four different empirical phenomena that involve conceptual enrichment in the integration of lexical and construction meaning. The constructions at issue are of varying kinds. Case studies 1 and 2 discuss argument structure constructions in the traditional sense, but from different ends: Case study 1 deals with constructions which express spatial relations and motion and explores their linking potential empirically, whereas case study 2 starts with the verbs and explores their combinatory potential with different syntactic and linking patterns. The case studies 3 and 4 explore constructions that can be treated as constructionist equivalents to derivational affixes in morphology, expressing a different meaning in case study 3 and a more grammatical feature, aspect or aktionsart, in case study 4. Each case study consists of corpus and experimental research. The aim of the corpus studies is to investigate the potential for variation in order to establish an empirical basis for the analysis of the enrichment phenomena involved in the four case studies.We will use the corpus resources that have been developed by the project B1, especially to investigate the role of constructions in spoken language and dialogue. The experiments will be carried out as web based experiments with the WebExp software package. Their purpose is to figure out the limits of conceptual enrichment. Which combinations of lexeme and construction can be accommodated by background knowledge or “saved” by context, and which cannot? The four experiments will each have the design of an acceptability rating experiment. The test sentences will be presented with and without linguistic context. Subjects will have to give an acceptability rating using a 7-point scale.