engleski

Gradient Grammaticality, Information Structure and the Parallel Architecture

In most contemporary formal models of grammar, it has often been assumed that there exists a so called "context-free grammar", that has "nothing to do" with how language is used. This is claimed to be the only way to objectively approach linguistics. As opposed to that, over the past decade or two it has become increasingly difficult not to observe that standard methods of judging grammaticality may be corrupted through estimation based on the frequency of usage, influence of the prescriptive norm, or specific lexical choices. This is true in particular when it comes to the existence of certain linguistic phenomena which obviously cannot be accounted for from the syntactocentric point of view, but they rather call to introduce a conceptual semantic tier into grammatical description. Among the most interesting phenomena that help us see the need to bridge the gap between phonological, syntactic and conceptual tier of grammatical description are clitic clusters in South Slavic languages. In order to be able to account for their role in the sentence phonology, as well as within the sentence information structure (IS), one needs to exercise an entirely new experimental approach. The starting point for this paper is twofold - one is methodological, the other one is theoretical. Methodologically, I distinguished between grammaticality, acceptability and acceptability judgments (Bard, Robertson, Sorrace, 1996). Theoretically, I follow the line of thinking that takes the information-structural properties to be part of grammar, and not extralinguistic (Lambrecht, 1994). By accepting these two premises, I reject the classical generative unidirectional input- output model of the relation between syntax and phonology, and take the IS to be the tier within semantics, which is connected both to syntax and phonology through interface rules. The main goal of my paper is to provide evidence that only an experimental research can resolve data disputes and, thus, open a path for a proper analysis of this phenomenon. In order to do this, experimental design will take into account he gradient data judgments in which the position of clitic cluster in Croatian should, according to syntactocentric theories, be accounted for as ungrammatical. For that reason, I am using magnitude estimation (ME). The second part of the paper will find a model of grammar in which the semantic combinatoriality that arises from the relative freedom of clitic cluster placement can be properly analyzed. To do that, I will use the framework of parallel architecture, and try to formulate interface rules that allow for this combinatoriality.

gradience; grammaticality; word order; parallel architecture

nije evidentirano