Natural languages exhibit all kinds of different instances of doubling: the phenomenon that a single semantic operator is manifested more than once in the morpho-syntax. In this talk I will discuss some doubling phenomena, such as negative concord, subject-verb agreement and modal concord.
I argue that these doubling phenomena behave all quite similar, therefore calling for a unified analysis. I argue that doubling can be best understood from a mismatch between semantic and phonological economy conditions. In short, the most economical way to derive a sentence with a particular semantic interpretation from the semantic (compositional) perspective is highly uneconomical from the other side of the spectrum given the phonological desideratum to express multiple markers of a semantic operator on one and the same lexical element (or structural node). In my talk, I demonstrate that such a phonology-biased way of marking semantic operators requires additional functional structure that can only be encoded by additional formal material that does not contribute to the semantics of the sentence. The manifestation of such redundant material is manifested as doubling.
The fact that different modes of marking semantic operators follows from different (conflicting) economy conditions, imposed by semantics, phonology or the lexicon makes it possible to explore the thesis that the grammatical variation is defined by the interaction of these economy conditions. If that is true, it follows that parametric variation is a derived, rather than innately present, notion.
Imperatives are ordinarily used to give commands (Kill Bill!), but sometimes they can also be used to convey permissions (Come earlier, if you like!). The challenge for a modal logic for imperatives is to reconcile these two contrasting usages. In my proposal, imperatives will be analysed as 'nabla' statements. The nabla is a modal operator that when applied to a set of propositions asserts that (i) each of these propositions is possible; and (ii) their union is necessary. The double nature of this operator, I will argue, is exactly what we need to reconcile command and permission uses of imperatives, and to account for the related phenomenon of free choice in imperatives (To continue, push any key!, Do your homework or help me in the kitchen!).
The acquisition of quantification has been a much debated subject. Whereas various authors claim that children do not show non-adult behavior in their interpretation of universally quantified sentences, others argue that children do. Recently, is has been suggested that the difference between weak and strong quantifiers provides the key for our understanding of children's non-adult behavior (Drozd and Van Loosbroek, 1999, Geurts, 2003, Hollebrandse and Smits, 2006). My PhD project explores this possibility of characterizing the acquisition of quantification in terms of weak and strong quantifiers. In this talk, I will introduce you to earlier work on the acquisition of weak and strong quantifiers and present new data regarding children's understanding of the quantifier 'many' (Smits, Roeper and Hollebrandse, in prep). Westerståhl (1985) observed that this quantifier allows a strong, proportional reading with respect to its second argument set. For this reading, different explanations are given in the literature, either in terms of cardinality (De Hoop and Sola, 1995), focus (Herburger, 1997, 2001) or relative proportionality (Cohen, 2001). Literature on the acquisition of quantification has lacked to address these semantic characteristics of 'many'. The experiment that I will present takes into account, methodologically and experimentally, the different semantic accounts that explain the different readings of 'many', identifies an appropriate semantic characterization of this quantifier for children and relates this to the acquisition of universal quantification. Data of 28 children of a truth-value judgment task containing sentences with 'many', 'many of' and 'all' show that they do not prefer a weak or a strong analysis of the quantifiers, but rather overgeneralize the use of Q-raising (a local process of raising that eliminates the difference between its first and second argument set, Herburger, 1997, 2001). This suggests that, in order to reach adult grammar, a child has to take a next step and distinguish between different ways of computing the relevant set comparing operations of weak and strong quantifiers (Q raising versus QR). I conclude that, only by taking such differences between weak and strong quantifiers into account, we are able to understand children's actual understanding of quantifiers and identify the possibility of characterizing the acquisition of quantification in terms of the syntactic, semantic and pragmatic features of weak and strong quantifiers.
This talk seeks to explore the possibilities of accounting for conversational implicatures based on optimal assertions in more generality than done in the paper that introduced the idea (Benz and van Rooij "Optimal Assertions and what they implicate", to appear). It will be argued that the set of situations where optimality of assertions licenses pragmatic enrichment are rather limited. In particular, the speaker has to be assumed competent in one way or another, as problems with misleading information show. If, on the other hand, the speaker is assumed competent, the predictions of the optimal assertions approach (for pure interpretation scenarios) prove to be equivalent to strong bidirectional optimality theory (BOT). In order to be able to account for synchronic partial blocking phenomena (Horn's Division of Pragmatic Labor) an extension of the notion of optimality is introduced, which is proven to be equivalent to weak BOT. It will be argued that the proposed extension is an intuitive implementation of forward induction, thereby finally providing the missing link between Dekker & van Rooij's (2000) non-standard game theoretic reformulation of BOT and game-theoretic tradition.
Scalar modifiers, like 'at least', 'at most', 'more than', 'or more' etc. have attracted some attention lately. For example, Geurts and Nouwen show that superlative scalar modifiers (SSM) differ from comparative ones (CSM) in several respects, which they trace to a modal component in the meaning of SSM, absent in CSM. What this and other relevant proposals do not address in a satisfactory way, is that SSM can't occur in the scope of sentential negation and antiadditive quantifiers like 'nobody' and 'never' while CSM can. I argue that this can't be derived from the modal component of SSI, because the modifier 'or more' shares the modal component with SSM, but isn't excluded from those negative environments. Not all downward entailing environments exclude SSM. They can occur, for example, in antecedents of conditinonals, as well as in the restrictor of universal quantifiers. Upon closer scrutiny, however, it turns out that there are cases where CSM can occur in these environments, where SSM can't. I will argue for a generalization of recent proposals by van Rooij and Schultz concerning exhaustivity and the meaning of 'only' to the meaning of SSM. This leads to a derivation of the polarity sensitivity of SSM, including their restricted appearances in antecedents of conditionals and restrictors of universals. Handling the relevant properties of SSM and CSM by means of general semantic/pragmatic mechanism can also explain the otherwise puzzling fact that the distinct properties of such modifiers discussed by Geurts and Nouwen and others are cross-linguistically robust.
Taking the cartographic perspective, the nominal and the verbal expressions all involve a number of functional projections directly predicted by their categorical nature. Some of these projections are shared (e.g. those related to topic and focus), some are parallel in the relevant respects (inner aspect and grammatical number), and some are specific for each category. In purely methodological terms, one would hope to establish a full correspondence between the functional skeletons of the two categories, and preserve only one distinction, that of the property determining the categorical status. In this paper, I make a step towards such a picture, arguing that VPs take count quantifiers and numerals just like NPs. The only difference is that most languages have no, or very few, lexical items with quantificational meanings that can appear in positions at the VP level. I discuss VPs involving quantified non-specific arguments, and argue that they involve a reading in which the quantifier is base-generated and interpreted at the VP level, but only lexicalized on the noun, as a kind of concord, or agreement. I establish a parallel between a certain type of the negative concord and quantifiers on non-specific arguments and propose that the mechanism responsible for this is a certain kind of feature valuation. When an argument lacks the specification of certain properties, such as grammatical number, quantification or specificity, it will inherit these properties from the closest c-commanding projection in which a corresponding property is specified. When the base-generated instance of the quantifier cannot be lexicalized, the structure is rescued by its lexicalization on the embedded argument, as a kind of agreement. The analysis introduces a new perspective of the relation between distributive and collective readings. I provide examples from Chinese, which illustrate the other, straightforward strategy: a direct quantification of eventualities by an overt quantifier or a numeral.
Modal verbs, by definition, allow for an epistemic reading, presenting an estimation of the likelihood that a proposition is true. However, this reading is not available in all tenses. In particular, the perfect form of a modal is incompatible with epistemic readings. This can, of course, only be demonstrated for languages in which modal verbs actually have a perfect form, as in Dutch Hij heeft ziek moeten/kunnen zijn ('he was obliged/able to be ill'), but it may also be observed for the English semi-modal have to (he has had to be ill): these sentences only allow for dynamic or deontic modal readings (ability or obligation), but not for epistemic ones.
In this talk, I will argue that this phenomenon is actually part of a much more general connection between aspect and modality. More specifically, I will try to provide a unified explanation for a range of puzzles that, in addition to the incompatibility of perfect and epistemic modality, includes the use of imperfective rather than perfective forms in counterfactual wishes (I wish I ?read/was reading/had read that book), the lack of an epistemic reading for perfective complements of modal verbs (you must read that book vs you must be reading that book), and the dominant presence of imperfective forms in (free) indirect speech contexts.
What all epistemic or, more generally, 'subjective' readings of imperfective forms have in common is the temporal ordering relations of simultaneity. Semantically, imperfective forms express simultaneity of a situation with an independently provided point of reference, which may be a point of perspective or an epistemic evaluation time (and/or the point of speech itself). Perfective forms and, in my analysis, this includes the participial part of the perfect construction - resist simultaneity readings of any kind and are therefore generally incompatible with epistemic readings.
If time permits, I will briefly go into the cross-linguistic validity of the imperfective-epistemic link, since counterexamples have been provided particularly from Slavic languages.
A central assumption within the field of first language acquisition is that children first learn to comprehend a given form before they are able to produce this form correctly. However, in certain cases children make errors in the comprehension of linguistic forms sometimes even years after they have shown mastery of the same forms in production. Such production/comprehension asymmetries, with correct production preceding correct comprehension, have been attested for basic word order, pronouns, indefinites, intonation, tense, implicatures, and modification. In this talk I will discuss the implications of these and other acquisition asymmetries for the nature and organization of the grammar.
In languages like Dutch and English, we use articles and plural morphology to set up referents in a conversational space (a student, books), and to refer back to them (the teacher, the students). Linguists have formulated correspondence rules stating that definite articles express determined reference (unique/anaphoric reference, Farkas 2002), indefinite articles introduce new discourse referents (Kamp and Reyle 1993), etc. However, we don't necessarily find articles and plural morphology in all languages. Many languages use bare nominals, i.e. noun phrases that lack an article or determiner (Chinese, Hindi, Slavic, etc.). In this paper, we argue that language variation in the expression of plurality, discourse reference, and determined reference arises from the interaction of faithfulness constraints governing article use and plural morphology with a general economy constraint blocking functional structure in the nominal domain (*FunctN). A series of rankings results in an OT typology ranging from Chinese (no singular/plural morphology, no articles), via Slavic (singular/plural morphology present, no articles), and Hebrew (singular/plural morphology present, definite, but no indefinite article use) to French (singular/plural morphology present, obligatory articule use with definite/indefinite distinction in all cases). The typology leaves room for exceptional uses of bare nominals in languages like English, Dutch, as in at school, mother and child, op kantoor, hij is advokaat.