Abstract This article provides a comprehensive survey on current research on information structure so as to clarify some ‘paradoxical’ effects stemming from the tension between data and theory. Paradoxes are here defined as unexpected data in light of certain assumptions held in mainstream literature. More specifically, we explore two possible sources of paradoxes: certain features of the experimental design and, above all, inadequate discourse models. Mainly considering dislocation and cleft sentences in French and Italian we suggest that some apparent paradoxes (such as non-focalizing clefts or dislocation expressing focus-related functions) can be conceived of as the effect of general pragmatic mechanisms and rhetorical strategies exploited by speakers. We also claim that these effects can be better understood through explicit models of discourse.

Standard theories of focus expressed by cleft structures, for instance (Beaver & Clark 2008; Krifka 2007), assume that the motivation for the use of focus is discourse relevance: focus establishes an answer to the question under discussion (Roberts 2004: 216). This account, however, lacks a theory of how alternative sets are generated in real discourse. We present a study in the non-cumulative self-paced reading moving-window paradigm that tackles this problem by measuring how manipulating the context preceding German es-clefts influences alternative generation. Our results illustrate the interplay between discourse relevance and individual relevance in this process, in that cleft sentences only receive a contrastive focal reading in contexts where the readers are in a position to previously generate an alternative set.

Unlike in Germanic languages, in the Romance family the prosodic realization of polarity focus is strongly restricted. Instead, we observe a wealth of formal means that involve other language domains, such as the lexicon and non-canonical syntax. In this article, we provide a fine-grained analysis of the contribution of marked syntax to the expression of polarity focus in French, Italian and Spanish, by examining two constructions: clitic dislocation and si che / si que. Our main goals are to understand the distribution of these structures and their cross-linguistic differences on the basis of data gathered from the Direct Europarl corpus. Based on a Question Under Discussion model, we identify the main contexts of use to which clitic dislocation and si che / si que appear to be a good fit thanks to their functional properties; in particular, we recognize the framing function of clitic left dislocation as one of the most relevant features justifying the affinity between these structures and the expression of polarity focus. Finally, we explain the profound asymmetry observed in the distribution of clitic dislocation and si che / si que in our dataset on the basis of inherent grammatical grounds (related to the absence vs. presence of a system of clitic subject pronouns in the three languages), as well as the typology of information structure within the Romance family, as put forward by Leonetti (cf. Leonetti 2010, 2014) in his recent works.

Frequency is defi ned in terms of number of occurrences of a given linguistic structure in a particular linguistic system or sub-system (as approximated by a suitable corpus). Frequency is assumed to be a possible determinant in usage-based models of language change, language acquisition and language processing. While the default assumption is that there is a non-trivial relation between frequency of occurrence thus defi ned and mental and structural representation, the frequency factor is being investigated with a view both to its explanatory potential and to its challenges in future research.

SummaryThis paper suggests a functional definition for each of the four central case categories (or the four central syntactic functions of the noun) in the Standard Average European (SAE) languages, i.e. the subject /nominative, the direct object /accusative, the indirect object /dative, and the adnominal complement /genitive. In contrast to traditional approaches, which are normally based on homogeneous sets of semantic case roles indifferently situated on the propositional level of the meaning of a sentence, this paper starts from a prototypicalist perspective and tries to relate the four case categories or syntactic functions to four fundamental communicative tasks that nouns may fulfil in an utterance : a discourse function (topic) for the subject /nominative, an interpersonal function (concerned person) for the indirect object /dative, a propositional function (immediate argument) for the direct object /accusative, and a referential function (basis of « apprehension ») for the adnominal NP /genitive. The classical semantic case roles, on the other hand, are regarded as semantic components of the verbal lexemes, and the apparent systematic correlation between certain case roles and certain syntactic functions /surface cases is seen merely as an effect of affinity between the semantic roles and the postulated communicative functions.