Formal systems for computing meanings often rely very strongly on syntactic structure, and rarely on other overt clues such as morphology. This makes them quite inflexible. The calculus of Referent Systems, due to Kees Vermeulen and further developed and implemented by Marcus Kracht, allows for much more flexible algorithms since it accesses both syntactic and morphological structure. The basic idea is that semantic composition is not performed by function application, but by operations which identify variables depending on the morphosyntactic properties associated with them. The main burden for calculating meanings of complex expressions is carried by the argument structure, which interfaces syntax, morphology and semantics.

Instead of analysing quantified NPs as denoting generalized quantifiers, I propose (within the framework of DRT) that (i) the semantic contribution of quantified NPs is similar to the semantic contribution of indefinite NPs and pronouns in DRT, namely to introduce of a (set) variable and some condition(s) on the variable, and that (ii) there are different ways of evaluating a relation between two or more sets of entities. Given these evaluation operations and that the introduced conditions (like the conditions introduced by indefinite NPs in DRT) do not refer to a scope set, we predict that the set of Bs which are not A is always irrelevant for the truth of Q A B (conservativity), a property which appears to have cross-linguistic validity. Further, since quantified NPs and indefinite NPs make a similar type of semantic contribution, this can be viewed as motivating their similar distributional patters. And finally, since the set variable can be assigned a semantic role independently of how the condition on the set is evaluated or satisfied, we don't run into the third problem.

While a considerable amount of work has focused on finding the most adequate restrictions on (i) the formal structure of natural languages and (ii) the relation between formal and semantic structure, less work has focused on clarifying how to restrict the semantic structure of natural languages. In order to know how to restrict semantic structure, it is necessary to know why one would want to restrict it. To sharpen this issue, I will discuss what is often taken to be one of the most important reasons for postulating compositionality, namely the explanation of successful communication by means of novel expressions. However, as it turns out, what the explanation of this requires is not actually compositionality itself (at least not as it is usually thought of), but a much more restrictive notion, which I formulated as hypothesis (H), namely that (i) linguistic structure is exhaustively characterised by means of a set of basic exponent-meaning pairs and a set of rules for combining exponent-meaning pairs, and that (ii) speakers and hearers use the same exponent-meaning pairs and the same rules for combining complex exponent-meaning pairs, albeit in different ways, when producing or respectively understanding a complex expression. After pointing out what I take to be some difficulties for two prominent semantic theories when viewed from the perspective of (H), I sketch some ideas for an alternative, in which (i) semantic operations access the dependencies of unsaturated entities directly by means of the dependency type, and not by means of the order in which the dependencies must be saturated, (ii) semantic composition can be formulated independently of the analysis of scope and binding, and (iii) no nonlocal rules are necessary in order to map the phenostructure of meaning into the tectostructure of meaning.

In this paper we discuss a number of languages with a multidimensional Differential Object Marking (DOM) system. In such languages overt object marking is determined by more than one argument feature. We will show that such argument features can be related to case marking in different ways. On the one hand, they can trigger the occurrence of overt marking, on the other they can be the result of it. We will demonstrate that different languages may prioritize the different argument features in different ways. These cross-linguistic patterns call for a more flexible approach to DOM than hitherto developed. We develop a sign-based declarative model that does not rely on hierarchies. Hierarchies are argued not to be necessary for language-specific descriptions but only as a comparative concept in cross-linguistic comparison.

In Mongolian the conditions under which the accusative on embedded subjects can be omitted are different from the conditions under which the accusative on direct objects can be omitted. On the one hand, with direct objects the accusative can only be omitted if the NP is an indefinite NP, whereas this is not the case for embedded subjects. On the other hand, the omission of the accusative on embedded subjects depends on the adjacency of this subject to the matrix subject and/or on which of the two NPs is more prominent, which is again not the case for direct objects. We conclude from this that the accusative marking on an NP does not indicate that this NP is a direct object, but that this NP is not the matrix subject. If this is on the right track, then in addition to the other functions case may have (see e.g. Butt (2006)), it can also be used to distinguish NPs across clause boundaries -- an unusual function of case. In order to explain why the accusative on embedded subjects of object clauses can be omitted, we proposed (i) a prominence principle according to which the most prominent NP in a sequence of NPs is the matrix subject and (ii) a difference in the status of case morphology and word order information about grammatical role assignment. The accusative principle overrides the prominence principle whereas the first argument principle conflicts with the prominence principle, explaining why the accusative cannot easily be omitted from the embedded subject, if it immediately follows the matrix subject and is more prominent than the matrix subject.

This paper is the first of two papers dealing with the analysis of the phonological, morpho-syntactic and semantic properties of clitic pronouns in Romanian. The aim of this paper is to present an integrated analysis of both the phonological as well as the morpho-syntactic properties of clitic pronouns. Based on this, I develop an analysis of the syntactic and semantic properties of the clitic-verb cluster (in particular an analysis of the syntactic and semantic properties of clitic doubling and clitic left dislocation), so that together these two papers provide a sign-based analysis of the phonological, morpho-syntactic and semantic properties of clitic pronouns in Romanian. In order to clarify the relation between clitic pronouns and full pronouns, I begin with a brief description of the personal pronouns in Romanian, in the course of which I clarify some terminology and the way I use it. Next, I present (as theory-neutrally as I can) the phonological and syntactic generalisations governing the realisation and distribution of clitic pronouns in Romanian, followed by a detailed review of one of the most detailed and comprehensive analyses of clitic pronouns in Romanian, namely the analysis provided in \citet{Monachesi:2005}. After pointing out what I take to be some important shortcomings of this analysis, I will present in an alternative sign-based analysis of clitic pronouns. The two main assumptions underlying my analysis are that (due to the sign-based character of the proposed analysis) the phonological rules operate in tandem with the syntactic (and semantic) rules, and that the syntactic rules operate on tuples of strings, not just on strings. The two most important consequences of these assumptions are (i) that we can pair general phonological rules with syntactic rules, resulting in a restriction of the application of the phonological rules to certain syntactic categories but not others, and (ii) that we can analyse the position of the clitic sequence relative to the verb independently of the order in which the clitics are combined with the verb. Finally, I sum up the main claims of the present analysis and then discuss how these claims relate to the (still ongoing) debate about the nature of what I have called `clitic pronouns'.

This paper is the second of two papers dealing with the analysis of the phonological, morpho-syntactic and semantic properties of clitic pronouns in Romanian. In The syntax of Romanian clitic pronouns I present an integrated analysis of the phonological and the morpho-syntactic properties of Romanian clitic pronouns up to the level of the verb cluster. The aim of the present paper is to complete the analysis of Romanian clitic pronouns by (i) analysing the syntactic properties of the clitic-verb cluster, and (ii) providing an analysis of the semantic properties of clitic pronouns. Together these two papers provide a sign-based analysis of the phonological, morpho-syntactic and semantic properties of clitic pronouns in Romanian.

In this paper I will present an analysis of the following two generalisations in SiSwati: (i) the subject marker is obligatory both with clause-external and clause- internal subjects, and (ii) the object marker is in complementary distribution with VP-internal object NPs, at least with verbs which are not marked for aspect. Section 3 will present the basic assumptions underlying the present analysis of these generalisations. The basic unit of linguistic knowledge is a sign consisting of a for mal and a conceptual unit, signs are combined by modes of combination, modes of combination strictly separate formal and semantic structure, and semantic functions distinguish predicate placeholders in terms of how these placeholders are construed. Section 4 contains an analysis of the two generalisations in terms of modes of combination. The two most important features of the analysis of these two generalisations is that (i) the analysis does not postulate that the subject marker has lost its semantic value when it co-occurs with the subject NP, and (ii) it does not conflate formal and conceptual structure. Section 5 contains discussion of the analysis of non-clefted questions in Chichewa proposed in Bresnan and Mchombo (1987), and a reformu- ˆ lation of the basic idea within the present framework, which avoids postulating two types of subject markers. Section 6 concludes.

In this paper I present an analysis of the conjunctive and disjunctive verb forms in Siswati, a Bantu language belonging to the Nguni group and which is spoken in Swaziland and Southafrica. First, I present a brief sketch of the basic clause structure in Siswati. Then I introduce the notions of conjunctive and disjunctive verb forms as they were used in Creissels (1996). Next I present the data and generalisations from Siswati, and finally I present a sign-based analysis of the conjunctive/disjunctive distinction.

It has sometimes been claimed that Uzbek has “no definite or indefinite articles” (Bodrogligeti 2001, p. 55). Contrary to these claims, we show that Uzbek has not just one but two indefinite articles, namely bir and bitta. Uzbek thus shares this crosslinguistically rather unusual property with languages such as Lakhota, which distinguishes between two “quasi-indefinite determiners” (Lyons 1999), Moroccan Arabic, which employs a “potential” and a “concretizing” (Harrell 1962), and Maori, a more well-known example of a language where we also find two indefinite articles (Bauer 1993). In section 2 we show that bir and bitta are indeed used as indefinite articles. First we present Heine’s theory of the development of indefinite articles from numerals, then we discuss the relevant aspects of the numeral and classifier system of Uzbek, and finally we provide textual evidence that both bir and bitta occur in various usages as indefinite articles. In section 3 we compare the occurences of these two indefinite articles in two novels, the first one published in 1926 and the second one published in 2001, and show that while bir has reached the last development stage already by 1926, the use of bitta is a more recent development (at least in written Uzbek). The second important conclusion from this corpus study is that the use of bitta appears to have spread to the expense of bir not just in written but also in colloquial Uzbek. Due to the inherent limitations of corpus studies, some important questions could not be answered by this methodology. In section 4 we present the results of an online questionnaire, which was designed to elicit acceptability judgements in order to address some of these questions. First, the absence of certain usages of bitta (from our corpus) is, of course, not evidence of the absence of this usage. Acceptability judgements can help settle this issue. Secondly, we wanted to know whether, and if so in which contexts of use, in modern spoken Uzbek the use of bitta is judged better than the use of bir, because if this were the case then this would corroborate the hypothesis that bitta has spread to the expense of bir. The main result is that when it is used as an indefinite article introducing human referents, bitta is at least as good as bir (except in predicative constructions), and indeed better if used to introduce referents whose identity neither the speaker nor the hearer knows or cares about.