Artemis Alexiadou (editor) - The Syntax of Argument Structure: Empirical Advancements and Theoretical Relevance

Bridging theoretical modelling and advanced empirical techniques is a central aim of current linguistic research (e.g. ).

show reflexes of core properties of verbal syntax in the processing of ergativity.

The present collective volume takes up the line of research represented in such studies. It originates from a workshop addressing the interplay between new empirical advancements and theoretical perspectives in the investigation of argument structure held on the occasion of the 38th annual meeting of the German Linguistic Society at the University of Konstanz. The volume contains six studies on the syntax of argument structure combining empirical and theoretical perspectives. The main focus lies on the relevance of empirical results achieved through up-to-date methodology for the theoretical analysis and modelling of argument-structure. The contributions tackle issues of argument structure from different perspectives addressing questions related to diverse verb types (unaccusative and unergative verbs, (di)transitive verbs, psychological verbs), morpho-syntactic operations (prefixation, simple vs. particle verbs), case distinctions (dative vs. accusative, case vs. prepositions), argument and voice alternations (dative vs. benefactive alternation, active vs. passive), word order alternations (object preposing) and the impact of animacy, agentivity, and eventivity on argument structure. They report data obtained through a variety of methods such as psycho- and neurolinguistics experiments, corpus studies and acceptability judgments.

It is by now commonly acknowledged that linguistic theories need to stand the test of the rapidly growing possibilities of objectivized quantitative empirical testing in order to be legitimate. The last decade has seen the rise of fields such as experimental syntax or experimental semantics and pragmatics, employing methods originating from psycho- and neurolinguistics such as acceptability judgments, reaction or reading time measurements, priming tasks, eye tracking, or event-related potential (ERP). The relevance of these new fields is visible in recent and upcoming book publications (Goodall ed. to appear; Sprouse ed. to ). These research directions involve a change in the focus of interest originally associated with experimental methods in linguistics. While from a psycho-linguistic and neuro-linguistic perspective the subject of interest lies on processes of language production and processing, here the focus shifts to questions of the adequacy of grammar theories and their predictions.

Another constantly growing field of empirical research in linguistics includes corpus linguistics, which provides the tools and techniques for analyzing ever larger amounts of data from natural language production. Coupled with a rapid development of statistical methods, as witnessed in recent years, these methods provide a powerful tool for the study of language in general and the testing of grammatical theories in particular. At the same time, the constantly expanding empirical methods not only make it possible to test predictions inherent in a theory or compare competing assumptions of alternative theories, but also create new types of data that demand corresponding theoretical developments that have the potential to lead to new, e.g. cognitively more adequate, linguistic theories (cf. Phillips to appear).

What are the challenges resulting from advanced empirical methods as employed in psycho- and neurolinguistics or corpus linguistics for theoretical modeling? Does the investigation of argument structure come with special prerequisites when it comes to empirical testing? How does empirical testing feed linguistic theory? We want to elaborate on two such challenges inherent in quantitative studies of argument structure: (a) gradience resulting from gradient acceptability or frequencies in observation data and forced choice acceptability studies; (b) disentangling effects of grammar and effects of processing.

The nature of the observed gradience immanent in quantitative data depends on the scale of the data. We obtain different numeric values for a given phenomenon, depending on the selected scale, e.g. in analyzing the proportions of aggregated data (by several speakers and/or several lexical items) or in calculating the central values of gradient acceptability judgments. A crucial question/controversy that has been addressed by many scholars using quantitative methods is the relation between gradience in experimental or corpus results and the nature of grammatical categories (e.g. realizations. Variation is expected in any type of repeated observations and it is taken for granted that different lexical realizations will not render the same result, even if the syntactic structure or operation at issue does not depend on the lexical realization. This is not evidence that the grammar itself is gradient, it is rather the null hypothesis in observing gradient data like proportions, time intervals, or acceptability values in a scale. At the same time, gradience in lexical categories may well reflect the residual of our understanding of grammar. The lexical items that we use in order to test the acceptability of a syntactic construction are not just idiosyncratic entries in a lexicon but finally multidimensional entities that can be further decomposed in semantic/syntactic features. Hence, suspecting lexical variation opens the challenge to identify and (empirically) establish the relevant semantic/syntactic features that may contribute to reduce the residual variation in the data.

Argument structure by its very nature seems to be more susceptible to or dependent on lexical variation since the lexical items themselves constitute argument structure classes. Quantitative studies on argument structure are commonly based on testing several lexical instantiations of a given class (as also evidenced in most studies of the present volume), e.g. they examine a number of verbs as representing a verb class. Here, gradience at the lexical level is particularly relevant since it is a major source of indeterminacy in establishing boundaries between verb classes. For instance, certain readings may not be categorically excluded for a verb or a class of verbs but may be possible in an appropriate context (see e.g. Darby et al. this volume on gradient agentivity and eventivity with psych verbs and a proposal of coercion being at play for a subclass of psych verbs). Furthermore, certain lexico-semantic properties of verbs may be less stable than others. For instance, ).