# Amsterdam Colloquium 2005

Amsterdam Colloquium 2005: List of Abstracts

# 2005 Vienna Circle Beth Lecture

### The Logical Structure of Cognition. Lessons from Carnap's Logical Structure of the World

Hannes Leitgeb

This lecture reconsiders Rudolf Carnap's classic Der logische Aufbau der Welt (The Logical Structure of the World) from a modern perspective and aims at bridging the gap between early Logical Positivism and more recent developments in the logical analysis of cognition. Although the underlying programme of the Aufbau has failed, there are still important lessons to be learnt from it - indeed we will try to show that parts of the original Aufbau can actually be saved. This is going to lead us to discuss various topics in logic, epistemology, and philosophy of science that are still as exciting and important as they were in the 1920s and 1930s: (i) methods of logical abstraction; (ii) the construction of concepts on the basis of subjective similarity; (iii) the dimensionality of conceptual spaces; (iv) disposition terms and the semantics of conditionals; (v) theoretical terms, the epsilon calculus, and the relativized a priori.

# Invited Talks

### A Second Time and Again

Sigrid Beck

This paper considers focus alternatives to presuppositional elements like again. We observe that there are empirical differences between again and its non-presuppositional counterpart a second time. A general question is raised about presuppositions in alternative sets.

### Proof-theoretic semantics for a syllogistic fragment

We present some prolegomena to Proof-Theoretic Semantics (PTS) for natural language (NL). The following quotation from Schroeder-Heister 2005 emphasizes the lack of applicability to NL, the original reason for PTS to start with: \begin{quote} Although the meaning as use'' approach has been quite prominent for half a century now and provided one of the cornerstones of philosophy of language, in particular of ordinary language philosophy, it has never become prevailing in the formal semantics of artificial and natural languages. In formal semantics, the denotational approach which starts with interpretations of singular terms and predicates, then fixes the meaning of sentences in terms of truth conditions, and finally defines logical consequence as truth preservation under all interpretations, has always dominated. \end{quote} In order to device a PTS for (a fragment of) NL, two steps are required:

* Device a proof-theory (a calculus) for the fragment, satisfying criteria proposed for PTS in logic. Replace truth condition by derivability conditions (in the above calculus) as the meaning of sentences in the fragment.
* Identify the contribution of subsentential phrases (down to words) to the PTS meaning of sentences in which they occur.

Here, we focus on the first task only.

### More on approximative number words

Manfred Krifka

The phenomenon to be explained in this talk is why round number words n measure expressions, such as one thousand kilometers, are interpreted in an approximate way, whereas non-round number words such as nine hundred sixty two kilometers are interpreted in a precise way. In a previous attempt to explain this phenomenon, I have assumed that speakers prefer short expressions and approximate interpretations. I will show here that it suffices to assume just the first preference; everything else follows from considerations of strategic interpretations. I will also discuss evidence for an evolutionary adaptation of scales of different granularity. Also, it will be shown that measurement scales provide good evidence that the expressions a language supplies to its speakers influence the way how facts about the world are represented.

### Natural Language, Natural Logic and Natural Deduction

Lawrence S. Moss

This talk is concerned with logical systems based on syllogistic reasoning. The overall question one should ask are whether proof theory could ever be as foundationally significant for semantics as model theory. I think the jury is still out on this question, but in order to give proof theory a chance I believe one should attempt to find logical systems for interesting linguistic phenomena which are complete. The matter is interesting also for computational semantics, and Nissim Francez' talk at our Colloquium will highlight this. There is not so much work on complete syllogistic fragments, and so I should be able to go into detail about much of the extant literature. I think that there are many interesting questions for both logic and semantics that come from the perspective of this talk, and my talk will mention those.

# Workshop on Language Learning

### Principles and Implementation of Inductive Parsing

James Cussens

I will discuss logical, statistical and practical aspects of the incremental induction of grammars from sentences. The logical framework is moulded on Shieber et al's "Principles and Implementation of Deductive Parsing". I propose that inductive logic programming has the potential (not yet realised!) to incorporate *semantic* constraints to help induction. As for statistics, I will consider whether recent work in what has become known as 'Statistical Relational Learning' is applicable.

### Locality and the order of acquisition steps

Jacqueline van Kampen

Preferably, the properties of grammar can be derived from the following factors: (i) The primary linguistic data as they are offered to the child. (ii) A language acquisition procedure. Hopefully, the language acquisition procedure will be compatible with plausible assumptions about the neural abilities of human beings, but that is of no immediate concern. The interaction of the primary data and the acquisition procedure can be studied by a closer look at the order of the child's acquisition steps. What does she acquire first and why? What does she acquire later and why? My main point will be that this is empirically a promising and by no means trivial approach. At the same time, I will argue against an assumption that is quite common in computational studies and also in mere grammatical studies of child language. People from Gold (1967) to Yang (2002) assume that the acquisition procedure has simultaneous access to all data at once. My point will rather be that the acquisition procedure implies a natural selection of data. The data selection procedure must predict the actual order of the acquisition steps in the various languages.

### Semantics-Driven Learning of Lexicalized Grammars

Makoto Kanazawa

I will give a rough outline of an algorithm for learning formal grammars (with both syntax and semantics) from data consisting of sentences (word strings) coupled with meaning representations. The grammar formalism is that of Abstract Categorial Grammars (aka Lambda Grammars), and the meaning representation language is typed lambda calculus. The learning procedure is semantics-driven in the sense that learning the meaning of a word and its type precedes learning its syntactic properties. Main tools come from proof theory and lambda calculus.

### Semantics learning from corpora and background knowledge

Claire Nedellec

The acquisition of semantic lexicon and ontolologies in specialized domains is costly because of the lack of human experts and linguistic resources. Machine learning from annotated and unannotated corpora can contribute to speed the process up. We will illustrate this issue by two examples: learning hierarchies of semantic classes and learning semantic relations from parsed corpora and background knowledge in the domain of genomics.

# Workshop on Semantic Universals

### Prohibitives: why most languages are not like Dutch

Johan van der Auwera

In Dutch prohibitive constructions use the imperative form of the verb and the maid of most negative work niet 'not'. Thus combining zing 'sing' and niet 'not' yields Zing niet! 'Don't sing!'. In human language, this is not the preferred strategy. The most striking feature of prohibitives—a statement based on an analysis of some 500 languages—is that possibly two thirds of the world's languages commonly use a negative marker that is more or less dedicated to the prohibitive use. The question is why this should be so, and also why one third of the world's languages can go against this preference. The explanation will be argued to be of a semantic-pragmatic nature, most importantly referring to the static nature of declarative negation, and the dynamic nature of prohibitive negation.

### Case and strength

Helen de Hoop

In the functional-typological literature two main functions of case-marking are distinguished. One motivation for case-marking is disambiguation, i.e. the need to distinguish between the arguments of a two- or three-place relation.  In order to differentiate the subject from the object it is not necessary to mark them both; a case marker on one of them serves to distinguish them. Another widely attested function of case involves the identification of specific semantic information by expressing it through case morphology. That does not only hold for lexical (or semantic) cases such as e.g., locative cases, but to a certain degree for structural cases as well. For example, dative case is associated with goal and experiencer semantics, ergative case is usually associated with agentivity, and accusative case is associated with patienthood. In this talk I will investigate the two strategies of case-marking to see where they converge and diverge with respect to the semantic features of the noun phrases that bear case. Crucially, the 'strength' of the case bearing arguments will be shown to be of utmost importance for case-marking. The 'strength' of arguments can be viewed as a function of their 'discourse prominence' or as their degree of 'prototypicality'. In general we will see that animacy and definiteness both contribute to the strength of grammatical arguments. The similarities between animacy and definiteness will be examined to get a clearer picture of the role of argument strength for case-marking. The aim of this talk is to explore the relation between the strength of nominal arguments and the 'meaning' of case.

### How Much Logic is Build into Natural Language?

Edward Keenan

First Order Logic (FOL) with equality is a universal grammar for a class of languages - Elementary Arithmetic, Euclidean Geometry, Set Theory, ... . It defines their syntax, semantics, and proofs. Learning a natural language (NL - English, Japanese, Swahili,...) entails *overt* learning of many first order structures: Function-Argument and Predicate-Argument expressions, Recursion, and Boolean Operations: finitary 'and', 'or', 'not' and unbounded 'all', 'some'. NL falls short of FOL in precision: it lacks the full equivalent of variable binding operators and it allows structural ambiguities ("John told Bill he was bleeding" - he = John?  Bill?  a third party?  "John didn't leave because the children were crying" (That wasn't why he left, or, that's why he didn't leave).  NLs exceed the expressive power of FOL with proportionality quantifiers (most, two out of three,...), cardinal comparison (more/fewer students than teachers signed the petition), and non-intersective adjectives (a tall student) as well as non-extensional expressions: 'too many', 'not enough'; 'skillful', 'good'.

Learning a NL also entails *covert* learning of logical notions as some grammaticality patterns are conditioned by logical properties of the items in the pattern. For example, which NPs license the presence of words like 'ever' and 'any' in Ss like: "No students / Fewer than five students here have ever been to Minsk", but ungrammatical is "Some students / More than four students here have ever been to Minsk".  Partial answer: NPs that denote monotone decreasing functions license 'ever', etc., monotone increasing ones do not.  Additional instances will be given in the lecture.

A last, deeper similarity to FOL:  a NL is a closure system - a set of "words" closed under certain structure building operations. The "structure" of an expression is what is invariant under the structure preserving maps (automorphisms) induced by these operations. Speakers "know" this structure (in the sense in which they "know" the language) and treat expressions with the same structure as meaningful "in the same way". So being meaningful in different ways implies difference in structure. Example the predicates 'good to eat' and 'reluctant to eat' in "That fish is good to eat" and "That child is reluctant to eat" are understood in different ways, so they are predicted to exhibit structural differences. For example 'good to eat' may frame a noun, as in 'a good fish to eat', but 'a reluctant child to eat' is not grammatical.  A deeper example concerns the reference of the pronouns in "The juror expected to punish him / himself" versus "the judge who the juror expected to punish him / himself". 'Himself' must refer to the juror in the first case and 'him' cannot. The opposite pattern obtains in the second case.

### Association: A Cross-Linguistic Experiment

David Gil

In a series of publications, I have argued that Riau Indonesian exhibits a number of syntactic and semantic features that characterize it as typologically exceptional.  The question arises whether Riau Indonesian is truly exceptional, or whether its apparently exceptional properties are a mere artefact of a particular descriptive approach.  In order to answer this question, it is necessary to compare Riau Indonesian to other languages through the same eyes, using the same objective and rigorous yardsticks.  This paper proposes one such yardstick, in the form of a psycholinguistic experiment designed to elicit truth-value judgments in different languages.

At the heart of the semantic analysis of Riau Indonesian is the claim that when two expressions X and Y with meanings P and Q respectively are combined, the meaning of the collocation X Y is derived from that of its constituent parts by means of the association operator, A (P, Q), which says that the meaning of X Y is associated in an unspecified way with the meanings of X and Y respectively.  For example, if ayam means 'chicken' and makan means 'eat', ayam makan means A  (  chicken,  eat ), or anything that has to do in some way with 'chicken' and with 'eat'.  In particular, the semantic representation A ( chicken, eat ) lacks any specification of thematic roles: the chicken could assume the role of agent, patient, or whatever might make sense in the context of the utterance.

This paper presents the results of an experiment designed to measure, objectively across a variety of languages, the availability of apparently associational interpretations: interpretations that appear to be obtainable from the association operator without reference to thematic roles or other semantic categories. Two kinds of apparently associational interpretations are sought:  (a) those in which what looks like a bare noun preceding a bare verb is interpreted as the patient (rather than the agent); and (b) those in which what looks like a bare noun in construction with a bare verb is interpreted as an oblique argument or even a non-argument (in the absence of prepositions or other such markings).  The experiment presents subjects with a sentence in the target language and two pictures; subjects are asked which of the two pictures is best described by the sentence. The experiment is ongoing; as of August 2005, over 1000 subjects in a dozen languages had been tested.

While non-isolating languages have near-zero availability of apparently associational interpretations, isolating SVO languages generally allow apparently associational interpretations to some extent, thereby setting such languages apart from most others.  However, amongst themselves, isolating SVO languages exhibit substantial cross-linguistic variation with respect to the availability of apparently associational interpretations.  In this regard, the position of Riau Indonesian amongst isolating SVO languages is not exceptional:  it falls in the mid range of Malayic languages, and in the mid-range of other isolating languages, in fact with substantially lower availability of apparently associational intepretations than other West Malayo-Polynesian languages such as Minangkabau and Sundanese.

# General Program

### Dynamic Situations: Accounting for Dowty's Inertia Notion Using Dynamic Semantics

Ido Ben-Zvi

A dynamic epistemic framework is provided, for dealing with common sense inferences based on partial information. It is claimed that such inferences make use of an extended context of 'relevancy' or 'salience'. The dynamic semantics framework is extended with situation structures based on this context. The progressive aspect is then interpreted as such an epistemic inference, where the inertia set is modeled as a set of situations which are minimal in a sense. Formal semantics are given and put to the test.

Keywords: dynamic semantics, situations, partial information, progressive aspect

### Exhaustivity, Homogeneity and Definiteness

Richard Breheny

In this paper, it will be argued that the Homogeneity Presupposition (Fodor 1970, von Fintel 1997, Beck 2001) does not provide an adequate account of the tendency of plurals to obtain exhaustive, 'any'-interpretations in negative contexts. We argue that Krifka's (1996) rule for plural predication would do better if it were somehow restricted to arguments which are definite. We suggest an analysis which locates the optionality in plural interpretations in definite noun phrases rather than the predication.

Keywords: Plurals, Homogeneity, Definiteness

### Comics Relief for Anankastic Conditionals

Tim Fernando

Purpose clauses implicated in the semantics of  anankastic conditionals are analyzed in an event  semantics where events are conceived as sequences  of snapshots -- that is comics. The account generalizes to other types of conditionals,  avoiding certain well-known problems that beset possible worlds treatments, such as logical omniscience.

Keywords: purpose clauses, anankastic conditionals, events

### Achieving Expressive Completeness and Computational Efficiency with Underspecified Scope Representations

Chris Fox, Shalom Lappin

Ebert (2005) points out that most current theories of underspecified semantic representation either suffer from expressive incompleteness or do not avoid generating the full set of possible scope readings in the course of disambiguation. In previous work we have presented  an account of underspecified scope representations within an intensional first-order property theory enriched with Curry Typing for natural language semantics. Here we show how filters applied to the underspecified scope terms of this theory permit both expressive completeness and the reduction of the search space of possible scope interpretations.

Keywords: intensional semantics, underspecified scope respresentations, expressive completeness, computational efficiency

### How to and how not to employ discourse relations to account for pseudo-imperatives

Michael Franke

Pseudo-imperatives (PIs) are conjoined sentences where an imperative clause is conjoined or disjoined with a declarative clause ("Do X and/or Y will happen/be the case/be done."). There is an intriguing pragmatic asymmetry between PIs with conjunction and PIs with disjunction which this paper elaborates. Recently, some authors (Lascarides & Asher 2004, Gomez-Txurruka 2002) have made use of discourse relations to account for the pragmatics of PIs. This paper rectifies these proposals and takes a stand towards general possibilities and limits of explanations based on ascriptions of discourse relations.

Keywords: pseudo-imperatives, discourse relations, multi-modal sentences, disjunction

### Agency and Case: A lattice-based framework

Scott Grimm

The typological literature has demonstrated that parameters such as agency, affectedness, and object individuation affect the realization of case-marking. The proposed analysis captures the specific contribution of such parameters, resulting in a model capable of explaining case alternations. A feature-based representation of agency properties is proposed, loosely based on Dowty's proto-role theory, but reformulated in terms of privative opposition and hierarchically organized via a lattice. Theoretical gains include wider empirical reach and greater simplicity, while practical results include a detailed analysis of the genitive/accusative alternation in Russian occurring with certain scope-ambiguous verbs, e.g. seek'.

Keywords: Semantics, Case, Agency, Thematic Roles, Lattice

### Dynamic Wh-Terms

Andreas Haida

The grammatical analysis of wh-questions in Groenendijk & Stokhof (1982) is unsatisfactory in that wh-terms are not treated in the same way as indefinites (although conceptually desirable and typologically suggested). In G&S (1992), it is pointed out that this unification can be achieved if existential quantification is dynamic. I will spell out this proposal: The question denotations of G&S (1982) are reproduced in a dynamic semantic framework in which wh-terms translate as existential GQs. The syntactic and semantic consequences for explaining various intervention effects are explored.

Keywords: question semantics, dynamic semantics, intervention effects, wh-syntax

### Contrastives and Gricean Principle

Yurie Hara

It has been observed that contrastive-markings in various langauages are associated with uncertainty implicatures. However, a sentence can be contrastive-marked even when the speaker has a complete answer to the question, as long as one of the alternatives have an opposit value from the rest. Following the analyses by Spector (2003) and  Schulz and van Rooij (2004) on exhaustivity and the Gricean Principle, this paper claims that Contrastive-marking presupposes that the speaker's knowledge is not maximal.

Keywords: Contrastive Topic, Implicature

### Inference, Ellipsis and Deaccenting

Daniel Hardt

While it has frequently been observed that inference is available for the interpretation of ellipsis and the licensing of deaccenting, it has not previously been observed that certain inferences are systematically unavailable, both for deaccenting and ellipsis. Sluicing is subject to a case-matching requirement, even in cases involving symmetric predicates, where an inference ought to allow case matching to be violated. Deaccented material is subject to a scope parallelism constraint, even in cases where the two possible scopes are inferentially related. I propose a theory of minimal inference, in which inference is restricted to a search among minimal submodels.

Keywords: Inference Ellipsis Sluicing Scope Semantics

### Asymmetries in language use reveal asymmetries in the grammar

Petra Hendriks, Helen de Hoop & Monique Lamers

How can it be that children know the linguistic rules necessary to produce a sentence, but show no knowledge of these rules when comprehending the same sentence? According to the classical view, the nature of the grammar is independent of its use. A puzzle for this view are cases where correct production precedes correct comprehension. For example, children who use grammatical function to determine word order in production not necessarily use word order to determine grammatical function in comprehension (Chapman & Miller, 1975). We show that the role of grammar in production is different from the role of grammar in comprehension. As a result, the grammar must be bidirectional.

Keywords: language acquisition, sentence processing, production/comprehension asymmetries, word order

### A Presuppositional Account of Indexicals

Julie Hunter & Nicholas Asher

Abstract: Many indexicals can pick out referents from a context other than the context of use, contrary to David Kaplan's predictions. English actual', here', and now', along with Amharic I' and Chinese ziji' can shift contexts. We propose that indexicals should be understood anaphorically. Indexicals trigger presuppositions which search for antecedents. Using DRT, we show that they search the global context first (the outermost DRS), but when binding in the global context leads to inconsistency, they will search within a local context. If binding fails entirely, accommodation will often produce an antecedent. This account is simpler than Philippe Schlenker's in that it relies entirely on existing mechanisms in dynamic semantics and DRT to explain the behavior of shifting indexicals.

Keywords: Indexicals, Actual, Kaplan, Presupposition

### IF logic as a strategic game

Theo M.V. Janssen

The traditional game interpretation of IF logic has sometimes been criticized. Here we propose an alternative: IF logic as a strategic game. The game is played by two teams, the A-team that tries to refute the formula, and the E-team that tries to confirm the formula. We base our semantics on two assumptions: (1) the players are 'rational': they do not play a strategy if there is a better one (2) the players know that the others are rational. A formula is true if there is a Nash-equilibrium with value 1 (true). In this semantics signalling is not possible. The semantics has consequences for the linguistic applications.

Keywords: IF logic, game theory, branching quantifiers, Nash equilibrium

### When Widening' is too Narrow

Jacques Jayez, Lucia Tovena

Current proposals that characterize the widening effect of F(ree) C(hoice) items as an implicature all require additional stipulations and leave a number of observations unexplained. We propose instead that Free-choiceness results  from ensuring that every member of the restriction CAN satisfy the scope but none MUST, via a constraint that applies to all FC items. Differences among items can then be traced back to (i) the interaction of the consequences of  this general instruction with the particular modal profile of each item (ii) its universal vs. existential nature and (iii) the defeasible or non-defeasible nature of its modal profile, thus keeping a strong unity for the class of FC items while making room for variation

Keywords: free-choice items, implicatures, equivalence

### Donald Duck is back, and he speaks Spanish

Luisa Martí

This abstract argues that the elegant solution to the "Donald Duck" problem put forth in Schwarzschild (2002) is, unfortunately, not viable. The argument is as follows. Schwarzschild's solution to the problem involves the idea that the domain restriction of indefinites can be a singleton set. This assumption not only solves the "Donald Duck" problem, it also explains why indefinites can take scope outside of syntactic islands in many languages. I show with data from Spanish, however, that there are indefinites whose wide scope is sensitive to islands. If wide scope readings are analyzed using the singleton-set idea, however, their sensitivity to the syntactic environment in Spanish is not expected. This suggests that we should reject the singletonness assumption. But if so, we no longer have a general solution to the "Donald Duck" problem.

Keywords: indefinites, syntactic islands, the "Donald Duck" problem

### The Compositional Semantics for locatives revisited

Cécile Meier

Spatial expressions are usually interpreted as relations between two individuals (see e.g. Bierwisch 1996). I am going to argue that a smoother picture of the semantics of locatives arises if we assume that locative prepositional phrases express properties of locations. This move necessitates the introduction of variables for locations into the formal language. I am going to show in detail that an analogous linguistic system underlies the reference to locations and the reference to times. Locative modifiers may play the role of frame-setters restricting the reference location. They may set the location of the speech or the location of an event, state or an individual (in analogy to the so-called event time and speech time). Furthermore, expressions as everywhere and nowhere act as locative quantifiers. In my view, the overall architecture of locative semantics mirrors the properties of other quantificational domains and this view fits nicely into the program of ontological symmetry that Philippe Schlenker recently started to develop.

Keywords: Locatives

### Comparatives without Degrees: A New Approach

Friederike Moltmann

In this talk I will pursue an approach to the semantics of comparatives without making central use of degrees, an approach that is based on the notion of a 'particularized property' or what philosophers nowadays call a trope.

Keywords: comparatives

### Synonymy, Common Knowledge, and the Social Construction of Meaning

Reinhard Muskens

In this paper it is shown how a formal theory of interpretation in Montague's style can be reconciled with a view on meaning as a social construct. We sketch a formal theory in which agents can have their own theory of interpretation and in which groups can have common theories of interpretation. Frege's problem how a proposition can be grasped is no longer solved by placing the proposition in some Platonic realm, but by making use of the common knowledge of language participants.

Keywords: Common Knowledge, Hyperintensionality

### Monotone Amazement

Rick Nouwen

This paper investigates how evaluative predicates (like 'amazing', 'surprising' etc.) combine with monotone and non-monotone statements. I argue that although these predicates give rise to different interpretations in different positions, they always licence the same polar orientation effects. Additionally, I argue that these effects are solely due to the monotonicity of evaluative predicates which preserves (or reverses) the entailments that are licensed by the structure to which the predicate applies. Finally, the different ways in which evaluative predicates convey their emotive content differ with respect to factivity. As I will show, adverbs like 'amazingly' can modify operators that convey a standard of comparison on the basis of intensions. When this is the case, factivity is lost.

Keywords: evaluative predication, monotonicity, polarity, vague quantification

### 'Almost': A test?

Doris Penka

Modifiability by 'almost' has often been used as a test for the quantificational nature of a DP. The aim of this paper is to give a semantics for 'almost' as a cross-categorial modifier. It is argued that 'almost' introduces alternatives in which the modified expression is replaced by a value close by on the corresponding Horn-scale. It is shown that such a semantics derives the correct truth conditions for 'almost' applying across categories and explains restrictions on its distribution. At the same time, taking the semantics of 'almost' seriously invalidates the 'almost'-test as diagnostic for the nature of quantifiers.

Keywords: semantics

### Determiners in aspectual composition

Christopher Piñón

A difficulty with leading theories of aspectual composition (Krifka 1992, Verkuyl 1993) is that they make incorrect predictions when verbs of gradual change combine with DPs containing determiners such as 'at least (three)', 'at most (three)', 'less than (three)', 'many', and the like. The problem is that such determiners form DPs that are not (in Krifka's terms) quantized and some of which are even cumulative, and yet they all yield accomplishments in aspectual composition, contrary to expectations. I discuss the details and review a solution to this problem due to Zucchi and White (1999), arguing that it is not convincing. I then propose a new approach to the constitution of accomplishments that appeals to the focus meaning (in the sense of Rooth 1992) of the determiners of internal argument DPs in aspectual composition. In brief, a VP or sentence may be an accomplishment either by virtue of its ordinary meaning being quantized or by virtue of a compatible alternative derived from the focus meaning of the determiner of the internal object DP that is quantized. I show that this approach is more successful than the others in accounting for the problematic data.

Keywords: determiner, accomplishment, aspectual composition

### Scope Disambiguation by Ellipsis and Focus without Scope Economy

Mats Rooth

This paper reanalyzes data on disambiguation of quantifier scope by ellipsis and focus which in Fox (1999) were attributed to an interaction between focus interpretation and a scope economy principle. It is shown that a more thorough development of the effect of focus structure allows the disambiguation to be analyzed purely in focus theory, without appeal to economy principles. The analysis has a simple formal character, in which scope representations are generated combinatorially, and filtered by focus constraints. As in Fox, the link to ellipsis is provided by the assumption that focus is involved in licensing ellipsis.

Keywords: Focus, Ellipsis, Quantifier Scope

### The helping-effect of dative case

Antonia Rothmayr

German 'helfen' (help) + DAT cannot be captured by standard applicative analyses. Employing a post-Davidsonian view, the paper derives the different stative/eventive readings of 'helfen'. Eventiveness is tied to DO and BECOME, but not to CAUSE. 'helfen' is related to other uses of dative in German via Brandt 2003.

### Against Partitioned Readings of Reciprocals

Sentences with reciprocal expressions can often have what we call a `partitioned'' interpretation. For example, the sentence 'The men are hitting each other' can be true in case there are different groups of men, and the hitting only goes within the groups, and not between men of different groups. In the literature on reciprocals, some works attribute such partitioning effects to a special reading of the reciprocal, while other works assume distribution of the antecedent set into disjoint subsets independently of the  reciprocal's meaning.  In this paper we study the availability of partitioned readings with definite subjects and proper name conjunctions. This shows new evidence that partitioned interpretations of simple reciprocal sentences are independent of the semantics of the reciprocal expression, and are exclusively determined by the interpretation of the subject.

Keywords: reciprocal expressions, partitions

### Syntax and Semantics of Causal denn in German

Tatjana Scheffler

This paper presents a new analysis of denn (because) in German. In addition to causal links between propositions, denn can express the causation of epistemically judged propositions or of speech acts. Denn's behavior is explained by two properties: On the semantics side, I show that denn is a conventional implicature item. Syntactically, denn is a coordinating conjunction of CPs. These facts explain two things. (1) Why denn can be used to express a wider range of causal relations than the related weil: denn can target the coerced variables over assertions as an argument, while these variables are too high for weil. (2) At the same time, the restrictions on the use of denn also follow from denn's status as a coordinating conjunction and conventional implicature.

Keywords: conventional implicature, connectives, relevance conditionals, because, German

### Transparency: An Incremental Theory of Presupposition Projection

Philippe Schlenker

Heim 1983 claimed that the Context Change Potential of a connective could be derived from its truth-conditional contribution. Following Soames 1989, however, this claim was retracted in Heim 1992: if the 'real' conjunction and has a CCP defined by C[p and q]=C[p][q], one can just as well define a 'deviant' conjunction and* which has the same truth-conditional contribution but the opposite presuppositional behavior (just set C[p and* q]=C[q][p]). Heim's theory was thus insufficiently explanatory and could not extend to connectives whose CCP was not known to begin with. We provide an alternative derivation of Heim's results which does not suffer from these drawbacks. The idea, which is implemented in a fully classical framework, is that a clause {p}p' with presupposition p and assertion p' must satisfy a condition of Transparency, which is checked as soon the clause is pronounced. Given a set of assumptions C, if {p}p* is pronounced after a string a, Transparency requires that, for any sentence completion b, and no matter what the content of p* is, p should be eliminable, in the sense that C |= for all p' (a [p & p'] b <=> a p' b).

Keywords: semantics pragmatics presupposition projection

### Exhaustive Imperatives

Magdalena Schwager

Imperatives are often taken to encode necessity. An ambiguity arising with 'for example' provides evidence in favour of interpreting imperatives as containing a possibility operator. A default process of exhaustification turns it into necessity operators, unless overtly blocked by 'for example'. 'for example' can also apply after exhaustification, thereby giving rise to a reading of non-exhaustive necessity. The observed ambiguity is accounted for in a compositional way.

Keywords: Imperatives Exhaustification Modality

### Word Meaning, Unification and Sentence-Internal Pragmatics

Important developments within Discourse Representation Theory in recent years include a more elaborate formalisation and account of presuppositional phenomena, as well as the integration into the theory of unification as a mode of composition. Focusing on these two issues, the following claims will be made: (i) the varying compositional impact of some adverbials, ranging from merely constraining the properties of a predicate to radically altering them, is most suitably modeled applying unification, and (ii) pragmatic mechanisms such as bridging, presupposition verification and accommodation can be plausibly applied also solely sentence-internally in describing the semantics of lexical items. To substantiate these claims, the analysis will centre around the German causal preposition 'durch' ('by', 'through', 'with'), with some excursions to other languages.

Keywords: Discourse Representation Theory, Unification, Causation, Presupposition, Accommodation, Bridging

### Connecting Causative Constructions and Aspectual meanings: A Case Study from Semitic Derivational Morphology

Reut Tsarfaty

This work aims at identifying aspectual properties of events denoted by morphological causatives in Modern Hebrew (MH). The main purpose of this investigation is to elucidate the kind of connection that can be drawn between causative constructions and aspectual meanings, two notions that are not so easily equated (Levin 2000). A secondary goal is to provide a further argument in favor of the systematic aspectual contribution of Semitic derivational morphology. Our theory is inspired by Smith's causal chain (Smith 1991) and builds on a thematic account of Semitic morphology presented by (Doron 2003). Combining a formal and empirical investigation we argue that the MH causative template shifts the viewpoint of the event onto its initiation and development phases, making it more appropriate for describing imperfective situations.

Keywords: Causative constructions, Aspect, Derivational morphology, Modern Hebrew

### Semantics for Possessive Determiners

Dag Westerståhl, Stanley Peters

We give a uniform account of a wide range of possessive determiners, including simple (John's), quantified (few doctors'), and partitive (each of most students'), focusing on certain (to our mind) important but frequently neglected features of their semantics. One is the mode of quantification over the 'possessed' objects: often universal, but other modes are allowed too. Another is what [Barker 1995] calls narrowing: we agree it belongs to the semantics of possessives but note a methodological problem that seems to ensue. A third is the role of definiteness for possessives: we compare in detail our account to the 'definiteness accounts' common in the literature. Fourth, we study the monotonicity behavior of possessives.

Keywords: possessives, generalized quantifiers, definites, monotonicity

### Scalar Use of 'Only' in Conditionals

Ventsislav Zhechev, Sveta Krasikova

We argue that sentences of the kind "You only have to go to the North End to get good cheese" are ambiguous and employ a scalar version of 'only' on one of their readings. So do the exceptive constructions – the cross-linguistic counterparts of 'only have to' sentences. 'Only' is treated as inducing a 'comparative possibility' scale on propositions. The properties of scale explain the absence of the prejacent presupposition that is usually associated with 'only'. The sufficiency meaning component is argued to be a pragmatic inference, not part of the truth conditions. We also discuss the selectional properties of scalar 'only' with respect to the embedded modal.

Keywords: sufficiency modals, scalar 'only', comparative possibility, pragmatic inference

# Alternates

### Focus and Negative Concord in Hungarian

Agnes Bende-Farkas

This paper presents a newly discovered exception to Negative Concord in Hungarian that does not involve double negation. Rather, it is like English it-clefts that contain two negative particles that correspond to two instances of negation, in two independent formulae. The paper extends a DRT-based analysis of Hungarian Focus to the simplest cases, mentions some implications for the division of labor between syntax and semantics and suggests a possible method of presupposition accommodation that is required by the more complex cases.

### Complex Anaphors – Ontology and Resolution

Manfred Consten and Mareile Knees

Complex anaphors are nominal expressions referring to propositional structured referents  (such as propositions, states, facts and events) while introducing them as unified entities into  a discourse representation. In our paper, we will describe anaphoric complexation processes  and their constraints in terms of ontological categories. Furthermore, we will provide a  resolution model for complex anaphors based on semantic as well as conceptual structures,  thus integrating DRT and cognitive approaches. An example of an ambiguous complex anaphor will be discussed in order to show the role of ontological constraints in complex  anaphora processing.

### Polarity Items in Before Clauses

Francesca Panzeri

The aim of this paper is to propose a re-formulation of the uniform definition  Beaver and Condoravdi (2003) proposed to account for the meaning of before and  after, such that it can account also for the polarity items licensing behavior of the two temporal connectives.

### The Role of Lists in a Categorial Analysis of Coordination

Michael Schiehlen

The paper proposes categorial analyses for coordination with multiple conjuncts, correlative coordination, and respectively coordination. It argues that in a categorial setting these phenomena can only be adequately analysed if a data structure of lists is introduced. To this purpose the Lambek Calculus is extended with the Kleene star, a connective that has already been explored in other substructural logics. Correspondingly, the calculus is extended with list-forming operators as motivated by the analysis of the coordination phenomena.