Abstracts

Invited Speakers

Alice ter Meulen(Groningen)
Three degrees of dynamic involvement: the case of temporal reasoning

The Tarskian scheme of classical logic requires a division of labour between syntactic construction and semantic evaluation, preceding any form of inference. DPL, DRT and FCS all seem to have aligned themselves squarely with this Tarskian tradition, admitting of the first degree of dynamic involvement by accounting for certain semantic relations in discourse. A second degree of dynamic involvement requires a dynamic logic of indexicals, which, if tense is indexical, is originally found in Kamp’s DRT account of temporal anaphora. But certain forms of temporal reasoning seem to suggest that an intentional blurring of the separation of interpretation and inference may make for a more natural dynamic system for natural language, constituting a third degree of dynamic involvement. In ter Meulen (1995) the main new features in its representational architecture of Dynamic Aspect Trees are constituted by three kinds of moves:

(1) spreading of static descriptive information (informational update),

(2) the creation of new structures, called chronoscopes (temporal update), and

(3) the movement of the point of update activity (perspectival update).

The representation of a past tense clause in temporal updates creates new temporal structure only when it refers to an event, as opposed to a state.

For events, i.e. transitions from beginning to end, the construction rule introduces a new node by simple dependent attachment, when the new information is consistent with the information in the current open node. When the current node is closed or when a local consistency check rejects the new information, the construction rule searches for the lowest node dominating the current node which contains still compatible information, attaching the new information to a newly introduced right sister node. This procedure amounts to a belief revision, seeking the minimal change guided by the architecture of the given DAT while adjusting the perspective on the information left in the former chronoscope.

This situated notion of inference is different from classical Tarskian deduction in several respects. Inference is now relative to a DAT constructed for the premises in sequence and to its current node, which must support the descriptive information of the conclusion. One and the same text may produce different DATs, admitting for different inferences out of which different meaning emerges. The dynamic operations in DATs are not triggered just by verbal inflections, but by composition with the aspectual properties of their arguments.

Different possible algorithms for running DATs should spell out how much informational update is computed where and when. This invites a division between low-complexity ‘inference’ with lower case ‘i’: local consistency management via the lexical knowledge base with some minimal logic, e.g. Horn-clause rules, and ‘Inference’ in upper case ‘I’: ternary structural rules for the above format. The actions of inference are dynamic and hence affect the architecture of the DATs. But the more powerful Inference engine verifies conclusions which have no dynamic effect, as they are supported as stickers on the current node of the DAT constructed for its premises. Dynamic inference may employ information from the external knowledge base as an oracle, or it may adduce lexical semantic relations to drive the inference. The original static meaning-postulates of Montague Grammar now are seen potentially to have dynamic effects in order to preserve the global consistency in a chronoscope.

A second way in which inferential actions may affect the DAT is in the spreading of static information represented in the stickers. Their ‘portability constraints’ are strongly reminiscent of structural logical rules, e.g. downwards spreading to all dependent nodes is the DAT analogue of equivocating a state with a set of instants between its onset and termination common in interval semantics. Informational updates may be coded in the labels attached to nodes in DATs using modal operators ranging over possible DAT-updates. E.g. PERF intro and PROG intro are cheap and quick ‘lower case’ inferences. Big Inference is still relative to a particular DAT constructed for the premises, its current node and its source node. Since the sticker logic by itself is static, structural rules like MON and CUT are easily seen to be valid.

MON If X, Y |- A then X, B, Y |- A only when B is a sticker

CUT If X, A, Y |- B and Z|-A, then X, Z, Y|- A only when Z is a sticker

In CUT A is already a conclusion, so it is itself a sticker. A new question arises which operations natural languages may have to preserve CUT. It is suggested that expressing information about the past with a present perfect is one such way to cancel whatever dynamic effects the corresponding past tense expression would have had. Caution should however be taken, for not all common structural rules are valid even in sticker logic. For instance, the order in which stickers are introduced into a DAT is instrumental in some forms of conditional temporal reasoning (see ter Meulen 1995, ch. 5). So stickers are not generally freely permuted as labels of nodes in a DAT. A much more restricted form of Permutation may still apply to allow variation in the order of stickers of the same kind, e.g. two adjacent perfective stickers are permutable as well as two adjacent progressive stickers.

This study teaches us about naturalizing the use of logic in systems simulating aspects of human reasoning. The structural display of information in the representational strategy of DATs provides not only a natural link to visualizing temporal reasoning, but regarding context structured as a DAT with a verifying embedding in event models may lead to a natural proof theory for human reasoning in time about time.

ter Meulen, A. (1995). Representing Time in Natural Language. The Dynamic Interpretation of Tense and Aspect in Natural Language. Cambridge: Bradford Books, MIT Press. (paperback ed. 1997)

 

Peter Gärdenfors (Lund)
Concept combination: a geometrical model

In the literature, there are many attempts to model concept combination. However, all of them seem to have problems with many everyday examples like "stone lion". In this paper, I will propose a geometrical model of concept combination that is based on conceptual spaces. I will first outline how regions of conceptual spaces can be used to model simple properties and concepts. The basic rule for the combination XY of two concepts X and Y that I propose is that the regions for the domains of the modifier X replace the values of the corresponding regions for Y. If the regions of X are compatible with the regions of Y, the result of combining X and Y can be described as the intersection of the concepts, as was proposed for the classical logical theory. Thus "green apple" denotes the concept where the color region of apple is replaced by the more specific green region. However, if the regions are incompatible, like in "pink elephant", the region for X overrules that of Y (which presumably is the grey region of the color domain). In such a case X revises Y and XY cannot be described in terms of intersections. In some cases, the regions of X may block some of the most prominent domains of Y, leading to rather drastic changes of the concept Y. For example in "stone lion", the representation of stone includes the property "non-living" which is presumed by many of the domains of lion. My proposal explains why the meaning of "stone lion" is an object made of stone that has the shape of a lion. By taking into account the effect of "contrast classes", the basic rule is amended to cover a broader class of examples like "red skin".

Manfred Krifka (Austin)
For a structured account of questions and answers

There are two important general frameworks for the contribution of focus information, Alternative Semantics, as developed by Rooth, and Structured Meanings, as developed by Jacobs, von Stechow, and others. They correspond to two important general frameworks for question meanings, namely proposition set theories as developed by Hamblin, Karttunen and Groenendijk & Stokhof, and functional or structured theories, as developed by Hull, Hausser, Ginzburg, and others. In this talk I will give arguments for a structured approach to both the meaning of questions and the corresponding focus in the meaning of answers. In particular, I will show that the proposition set analysis has problems with predicting the right focus structure in answers (it cannot exclude over- or underfocused answers), that it is unable to distinguish between polarity (yes/no) and alternative questions, and that it does not allow to formulate an important condition for a type of multiple questions. On the other hand, I will show that one argument against the structured meaning framework, namely that it does not give us an elegant way to account for embedded questions, does not withstand closer scrutiny.

Contributed Talks

Dorit Abusch and Mats Rooth (Stuttgart)
Empty-domain effects for presuppositional and non-presuppositional determiners

Consider a speaker B for whom it is a stable and uncontroversial fact that the United States has no nobility. For B, sentence (1) is odd, to the extent that one might feel motivated to analyze it as truth value-less relative to B’s beliefs.

(1) Every American prince was at the party.

Sentences (2a,b) with cardinal determiners can be viewed as odd in the same way.

(2)
a. Two American princes were at the party.
b. No American princes were at the party.
c. Certainly not, only women attended.
d. Certainly, only women attended.

Yet as pointed out by Lappin and Reinhart (1988) and Reinhart (1998), B can reject (2a) as in (2c) and accept (2b) as in (2d). Our basic hypothesis is that when (1), (2a) or (2b) is viewed as odd relative to B’s beliefs, it is because truth is resolved by facts which are considered uncontroversial. On the other hand, (2a) can be rejected and (2b) can be accepted by evaluating them relative to a limited set of facts about the party, ignoring the stable global fact which makes them nonsensical. Sentence (2) can not be rejected in the same way, because of non-monotonicity in the truth value of every-sentences.

The analysis is extended to discourses where sentences with uncontroversially presuppositional determiners such as both are rejected in the same way. In light of the analysis we reconsider the question of the presuppositionality of the determiners every and most.

Maria Aloni (Amsterdam)
Interrogations under conceptual covers

I will present a refinement of the update semantics for questions presented in Groenendijk (1998), which involves relativizing questions to specific conceptualizations of the universe of discourse. In update semantics, who-questions partition information states. In particular, a question of the form ‘Who is P?’ reorganizes the structure of the state, so that only possibilities in which the extension of P is the same are grouped together. I will argue that this standard analysis of who-questions leads to a series of difficulties in connection with the notion of logical answerhood: Either our notion fails to capture the typicality of question-answer pairs like ‘Who is white?-Socrates is white’; or our theory fails to account for the non-triviality of identity questions like ‘Who is Socrates?’.

I will propose a modified analysis that avoids these problems. The main idea is that what counts as an answer to who-questions depends on the conceptualizations of the domain of individuals which are active in the specific circumstances of utterance. Formally, questions will be relativized to contextually given conceptual covers. Conceptual covers are sets of individual concepts that supply a perspective over the universe of discourse. Questions of the form ‘Who is P?’ will disconnect those possibilities in which the extension of P is ‘seen’ by different sets of concepts in the active conceptualization.

Renate Bartsch (Amsterdam)
The formal relationship between dynamic conceptual semantics and connectionist modeling

In this talk a mapping shall be stated between the level of experiential concepts in Dynamic Conceptual Semantics (DCS, Bartsch 1998) and concept formation in distributively representing connectionist neural models (CM). Both are formal approaches to learning linguistically expressed concepts, and at the same time they imply theories of understanding situations in perception and of understanding simple sentences that describe situations. DCS gives a logico-philosophical reconstruction of concept formation and understanding on the cognitive level of conscious phenomena, while CM models on an abstract level of neuronal connections and activation patterns the learning of classifications and associative connections in perception and behavior, also linguistic behavior, as generalizations over sequences of examples.

Concept formation in CM shares with DCS certain welcome properties: content addressability, stabilization, flexibility, and conceptual dynamics. Semantic systematicity can be captured in both models by part-whole relationships, unions, and overlap between activation patterns in CM and between sets of data in DCS. The mapping maps intersections in DCS on unions in CM and reverses inclusions. Overlap between activation patterns has to be taken into account as additional strengthening of activation, which will be shown to be indicative for syntactic-semantic structuring, i.e. the systematicity in distinguishing the semantically effective representations of, for example, Peter beats Paul from Paul beats Peter. >From the semantic properties explicated in DCS, which are based on principles of interpretation and truth evaluation, conditions are derived constraining the structures which stabilized neuronal networks must exhibit, will they be semantically effective.

In the talk I shall concentrate on the experiential level of concept formation, because there we can formulate a relationship between connectionist neural network models and the data-based dynamic model of concept formation. Naturally, connectionism cannot say much about theory formation, and therefore theoretically and formally explicated concepts will not be treated here.

Bartsch, R. (1998). Dynamic Conceptual Semantics: A Logico-Philosophical Investigation into Concept Formation and Understanding. SiLLI-series Stanford. Distribution: Cambridge University Press.

Fodor. J. (1997). Connectionism and the problem of systematicity (continued): Why Smolensky’s solution still doesn’t work. Cognition 62, pp. 109-119.

Jeannette Chur (Tübingen)
Sekundäre Prädikation bei Accomplishments: keine Unmöglichkeit

Gängige Meinung ist, da sekundäre Prädikation und Transitionen wie Achievements und Accomplishments nicht miteinander kompatibel sind. Für scheinbare Gegenbeispiele werden zwei Strategien vorgeschlagen: Entweder werden die Transitionen als Prozesse uminterpretiert, oder aber die sekundäre Prädikation wird zu einer Modifikation erklärt. Es soll gezeigt werden, da es mindestens einen Fall gibt, der dadurch nicht ausreichend erklärt werden kann: effizierende Verben mit einem Direktional als sekundäre Prädikation. Zentral wird dabei das Verb machen behandelt, welches zwischen overter zweiter Prädikation (mit Argumentstatus) und sekundärer Prädikation (mit Adjunktstatus) differenziert und somit zu ansonsten nicht erklärbaren Ambiguitäten führt.

Die Analyse führt zu folgenden generellen Ergebnissen:

  1. eine Differenzierung der Verben in effizierende, pseudoeffizierende (schieen) und nichteffizierende manipulative Verben (legen),
  2. Gleichzeitigkeit als Basis für resultative Konstruktionen und somit eine einheitliche Zurückführung aller sekundären Prädikationen (depiktiv und resultativ) auf den Mechanismus der konjunktiven Adjunktion verbunden mit
  3. Differenzierung der sekundären Konstruktionen auf der Basis des Ereignistyps der sekundären Prädikation, und
  4. Zurückführung der Restriktionen bei resultativen Konstruktionen auf generelle konzeptuelle Gesetzmäßigkeiten statt auf den semantischen Typ des primären Verbs.

Bernard Comrie (Leipzig)
Meaning and translation equivalence: How universal is the semantics of relative clauses?

A major concern in the narrowing down of the semantics of a construction is to determine the extent to which the interpretation of that construction is a function strictly of its semantics, versus the extent to which this interpretation is the result of other factors, including general inferencing strategies. Given this, it is quite possible that a sentence in each of two languages might constitute a pair of translation equivalents, but that they might differ in meaning, since aspects of the interpretation that in the one language follow from the semantics of the construction might in the other be derived rather from inferencing. In this paper, I aim to investigate this phenomenon by looking specifically at the phenomenon of relative clauses from a typological, cross-linguistic perspective. More specifically, I wish to investigate whether syntactic variation across translation equivalents of relative clauses necessarily involves semantic equivalence, or whether this functional equivalence transcends the boundaries of constructions identifiable as semantically equivalent.

For instance, in many Australian languages the same construction, with an adjoined dependent clause, serves as translation equivalent of both English relative clauses and various English adverbial subordinate clauses. In a number of languages, including Japanese, the same construction translates not only English relative clauses but also fact-S constructions and others. I develop an argument that such constructions in Australian languages and Japanese should receive a uniform syntactic and a uniform semantic structure, other aspects of interpretation following from inferencing.

Paul Dekker (Amsterdam)
Dynamics and pragmatics of "Peirce's puzzle"

It has been argued recently that a puzzle due to Charles Sanders Peirce (Peirce (1906)) crucially involves un-bound, or E-type, pronouns, and that, therefore, a dynamic or E-type analysis of the puzzle should be given. In this paper we will argue that "Peirce’s Puzzle" doesn’t involve un-bound pronouns so that no dynamic or E-type analysis of pronouns is called for, and that Peirce’s own solution gives intuitively right results, but in an arguably ad hoc manner. Next we will show that Peirce’s results can be seen to follow from independently motivated principles of dynamic semantics, viz., pragmatic / semantic principles governing the use of indefinite noun phrases.

Assinja Demjjanow (Berlin)
A semantic analysis of the Russian perfective aspect

I will show that it is possible to separate aspectual from other kinds of information in verbal prefixation in Russian, and to systematically relate them via semantically characterizable verb classes. A contradiction test can be formulated for all the meanings of prefixed verbs in, which shows that simple prefixation introduces a characterization of some state.

(1)
On bezal po mostu ceres reku, no ne dobezhal do drugogo berega.
He was running over the bridge across the river, but he did not reach the other shore.

(2)
On perebezal po mostu ceres reku, no ne dobezhal do drugogo berega.
He pere-ran over the bridge across the river, but he did not reach the other shore.

The proposal is that the relation e "abut" s or s "abut" e, which is characteristic of the perfective, is interpreted in Russian not only as immediate temporal precedence, but as weak contingency in addition. Each prefix chooses one of the conditions in e "abut" s or s "abut" e, or both. But it must specify different post- (or pre-) states for different verbs.

I propose that this specification is relative to a limited number of verb classes characterized by properties like ‘movement’, ‘bringing something into existence’, ‘having the terminal point defined by some norm’, etc. The knowledge about the post-states of the verb classes in the form of axioms constitutes a partial definition of linguistically relevant aspects of e "abut" s. It is therefore available for all prefixes. Thus, in all prefixes requiring a post-state, the post-state s of a movement verb like bezhat' is characterized as ‘the location of the moving object is at the end of the path’. The contradiction in (2) can therefore be uniformly reconstructed for any prefix using e "abut" s. The talk will be held in German.

Regine Eckardt (Konstanz)
Meaning change in truth value based semantics

In my talk, I want to outline an integrated theory of synchronic semantics and dia-chronic meaning change, based on truth value semantics in the spirit of Montague and his followers. In order to explain meaning change, we need to explicate the way in which words and their meaning are rooted in the speaker community. I will take causal theories of reference as my starting point in this question. However, these theories being designed for the case of proper names exclusively, I will argue that certain generalizations of later versions of the theory (e.g. Sterelny (1983)) are nec-essary: First, the range of possible denotata will be extended from rigid denotations to real intensional objects. This will strengthen, secondly, another parameter of bap-tizing situations which has so far held a marginal position: the answer to the 'qua'-question. To put it simply, if the meaning of a word is to be fixed on the basis of os-tensive reference to a single object, one needs to know in what respect this object is a sample for a larger class. The RESPECT IN WHICH we are to see the object an-swers the 'qua'-question. The resulting dynamic causal theory of reference will em-body a considerable potential for meaning variation. We will discuss the emergence of new "readings" of old words, as soon as enough ostensive contexts of use support a different answer to the 'qua'-question than the one which led to reading no.1. One example will be the re-interpretation of horn (=part of an animal) to horn (=musical instrument, originally made of horn no.1).

Sterelny, K. (1983): Natural kind terms. Pacific Philosophical Quarterly 64, pp. 110-125.

Ch. Ebert, D. Glatz, R. Klabunde & R. Porzel (Heidelberg)
Semantische Repräsentationen für die Produktion von Fokus/Hintergrund-Gliederungen

Während eine Vielzahl von Einzelanalysen zur (modelltheoretischen) Interpretation der Fokus/Hintergrund-Gliederung (FHG) existieren, sind die semantischen Bedingungen für die Produktion von Äußerungen mit einer FHG bislang nur in wenigen Arbeiten thematisiert worden. In unserem Beitrag wollen wir anhand spontansprachlicher Raumbeschreibungen zeigen, daß bestimmte vereinfachte Bedingungen, von denen für die Interpretation ausgegangen wird, wie z.B. die Simulation des Kontexts durch eine explizite Frage oder die Ausblendung von Planungsfehlern, für die Produktionsperspektive nicht angemessen sind. Wir präsentieren ein Modell, in dem auf der Basis der kommunikativen Aufgabe, die ein Sprecher mit seiner Äußerung lösen will, eine Default-Verteilung von Fokus und Hintergrund geplant wird. Diese dient als semantische Eingabe für eine HPSG zur Generierung von Einzelsätzen und zur Bestimmung des Fokusexponenten. Die Idee ist dabei, für jeden Fokus die kleinste Phrase mit einem Exponenten zu finden.

Markus Egg (Saarbrücken)
Reinterpretation phenomena in underspecified descriptions

The aim of this paper is to show that the method of accounting for semantic ambiguity by underspecified descriptions makes possible an intuitive and transparent treatment of reinterpretation phenomena as they are expounded by Moens and Steedman, Dölling, and Pustejovsky.

Reinterpretation avoids impending conflicts (e.g., of type or of sort) between the semantic contributions of the parts of a complex constituent by deriving additional information that allows a smooth composition of these otherwise incompatible contributions.

The high frequency of reinterpretation phenomena and the comparative ease of producing and processing them suggest that they should be accounted for by a semantic construction not too different from the one of non-reinterpretation cases. What is more, reinterpretations crucially introduce ambiguities on the semantic level, which can only be resolved with the help of contextual and world knowledge.

I analyze reinterpretation as an instance of the general phenomenon of semantic ambiguity. The analysis is part of a larger enterprise to describe ambiguity by underspecification. Ambiguous expressions are represented in a semantic metalanguage in a compact, non-disjunctive way by an underspecified representation. Each of its feasible specifications denotes an object-language reading of the expression.

So far, research within this framework has mainly focused on ambiguities arising from quantifier scope and ellipsis phenomena. Extending this framework to reinterpretation phenomena allows a straightforward account of semantic composition even in the case of reinterpretation and draws a firm line between the semantic and the non-semantic aspects of reinterpretation.

Jan van Eijck (Amsterdam)
Formal concept analysis and lexical semantics

To ascertain that a formalization of the intuitive notion of a `concept' is linguistically interesting, one has to check whether it allows to get a grip on distinctions and notions from lexical semantics. Prime candidates are notions like `prototype', `stereotypical attribute', `essential attribute versus accidental attribute', `intension versus extension'. We will argue that although the current paradigm of formal concept analysis as an application of lattice theory is not rich enough for an analysis of these notions, a lattice theoretical approach to concepts is a suitable starting point for formalizing them.

A well known way to represent a concept formally is as a pair consisting of a set of objects X and a set of attributes Y, linked together by the requirements that the objects determine the attributes and vice versa. The standard lattice theoretical treatment of this is called formal concept analysis (developed by R. Wille cs. in Darmstadt). Applications of concept lattices have mainly been in the area of classification theory and databases. In the talk I will show how concept lattices can form the starting point for a theory that is also linguistically interesting.

Stefan Engelberg (Wuppertal)
The "Imperfective-Paradox"-Paradox

In contrast to common assumptions, accomplishment verbs don’t show uniform behavior with respect to the so-called imperfective paradox, since some accomplishment verbs in progressive form imply a resulting state. The sentence ??Rebecca was killing Jamaal, but in the end, he wasn’t dead involves a certain kind of contradiction, as opposed to Linda was drying her hair, but in the end it wasn’t completely dry. This is the "imperfective-paradox"-paradox. The explanation of this phenomenon requires the distinction of two types of accomplishment verbs. The meaning postulates related to to kill express that to kill implies a resulting state, while to dry carries no such implication. (That to dry in simple form nevertheless shows up as an accomplishment is guaranteed by a lexical meaning postulate which states that a process of drying can not proceed infinitely without culminating.)

The resulting state (Jamaal being dead) in the first example sentence with the progressive of to kill cannot easily be negated. This is possible under certain circumstances, though: i) the speaker doesn’t know if Jamaal really died, but he assumes that this must have been the unavoidable consequence of what he saw: What was happening at five o’clock?-She was killing Jamaal–as far as I could see. ii) An intervening event is mentioned: Why did you shoot Rebecca?-I had to do something. I mean: she was killing Jamaal!. There are no such restrictions for the progressive of to dry. Neither do we have to look into the future to claim that she was drying her hair, nor do we have to mention an intervening event in case that the hair didn’t dry completely. This reveals a striking parallel between the meaning of the progressive aspect and the future tense: The meaning of a sentence like Rebecca is killing Jamaal in present progressive form (‘Rebecca is doing something, such that Jamaal will be dead afterwards, as far as the speaker can foresee and if no intervening event occurs’) is basically the same as the corresponding sentence in simple future tense: Rebecca will kill Jamaal (‘Rebecca is doing something / did something (?) / will do something such that Jamaal will be dead afterwards, as far as the speaker can foresee and if no intervening event occurs’). Thus, the assumption about the meaning of the progressive is the following: A verbal predication in progressive form with respect to reference time t claims that the lexically specified subevents en whose event times include t do in fact occur at t, while the subevents which are lexically specified as following en will occur in the future of t, where "in the future of t" has the modal-temporal meaning of the future operator relative to t.

Silvia Gennari (Providence)
Propositional attitudes and indexical tenses

Indexical tenses such as present tense occurring in intensional domains display a mismatch between the actual content of the propositional attitude and the attitude reported. Current solution to this problem propose that there is a de re interpretation of the embedded tense, parallel to those found with regular NPs. In this paper, I argue that this solution is not entirely satisfactory because of two reasons: First, the de re interpretation does not make the correct predictions for cases in which the content-report mismatch obtains but the interpretation has the properties found in de dicto interpretations. Second, the interpretation of tenses themselves do not seem to have the properties typically found in de re interpretation of NPs. As an alternative, I propose that reports containing indexical tenses involve an attribution of an implicit attitude similar to those traditionally acknowledged in the philosophical literature. I argue that such attributions are felicitous if there is an inference attainable in the common ground that allows the speaker to infer and assert the attributed content. The inference involves the hypothetical explicit content attributed to the attitude holder and pragmatic premises normally taken for granted. This account correctly predicts whether a given report is felicitous on the basis of the availability of the inference.

Bart Geurts (Nijmegen/Osnabrück)
Negative quantifiers and complement anaphora

The interpretation of a strong quantifier involves two sets: a domain of individuals, which is presupposed, and a subset of this domain (the ‘reference set’). Once these two objects are given it is possible to derive a third object, which is the set-theoretic difference between the two, but there are reasons to assume that this complement set is less salient. Moxey & Sanford (1993) (=M & S) argue however, that this does not hold for all quantifiers, because monotone decreasing quantifiers license, and even favor, anaphoric reference to the complement set. Their claim is based on an experiment in which subjects were asked to continue text fragments like the following:

(1) {Not many/Few/A few/...} MPs were at the meeting{. They .../because they ...}

The outcome of this experiment was that, whereas monotone increasing a few induces subjects to refer back to the reference set (i.e. the set of MPs that were at the meeting), monotone decreasing quantifiers like not many and few bias them towards the complement set (i.e. the set of MPs that were not at the meeting).

In my talk I will criticize M & S’s experiment as well as the conclusions they draw from their findings. I concede that M & S have shown that complement anaphora exists, and is sometimes even preferred. But it doesn’t follow from this that monotone decreasing quantifiers focus on complement sets. M & S’s results are caused by a conspiracy of factors, which create a somewhat artificial situation in which subjects are more or less forced to produce complement anaphora. One of these factors may even be the language: an informal replication of M & S’s experiment with native speakers of Dutch suggests that in Dutch the preference for complement anaphora is much weaker than it is in English.

Moxey, L. M. & A. J. Sanford (1993). Communicating Quantities. Hove/Hillsdale: Erlbaum.

Brendan S. Gillon (Montreal)
The mass count distinction: specification or underspecification

Virtually every formal semantic treatment of common nouns assumes that mass nouns and count nouns have different kinds of denotations. In particular, such treatments adopt the axiom of the Homogeneity of Reference, which holds that, if a mass noun is true of something, it is true of each of its parts. (See, for example, ter Meulen (1981), Link (1983), Roeper (1983), Lönning (1987), Eschenbach (1992), and Ojeda (1993), to mention only a few.) For many languages, including English and French, this axiom is refuted by a host of minimal pairs such as carpets versus carpeting or suggestions versus advice.

The paper presents a semantics for English mass nouns based on under-specification, a property exhibited in mathematical notation. For example, set theoretic notation uses upper case letters to denote or range over just sets and lower case letters to denote or range over both sets and non-sets. The mass count distinction is perfectly analogous: Just as upper case letters denote or range over just sets, so the denotations of count nouns are atomic and nothing else; and just as lower case letters denote or range over anything, including sets, so the denotations of mass nouns may either be atomic or non-atomic.

All this can be captured with two bivalent features, as will be shown in the presentation. Time permitting, it will also be shown how this analysis extends to French and Mandarin Chinese.

Jeroen Groenendijk (Amsterdam)
The logic of interrogations

I will present an update semantics for a language of predicate logic enriched with simplex interrogatives. The dynamic interpretation will guide us to new logical territories. Although under the classical interpretation A |= B? provides us with a reasonable notion of logical answerhood, under the dynamic interpretation of the language, it becomes salient that what A |= B? really means is that B? is a silly thing to ask if you have just been told that A. That raises the issue what, then, IS an appropriate notion of answerhood in a dynamic setting? The answer lies in the fact, that the update interpretation for the language gives rise to a new logical notion, next to and at the same level of generality as the notion of entailment. The new notion, which we call licensing, concerns logical relatedness of a sentence to the preceding context. An appropriate logical notion of answerhood comes out as the special case of A being contextually related to B?, i.e., B? licenses A.

The new notion of answerhood we arrive at, is both less and more demanding than the classical one. It is less demanding in that it characterizes partial answerhood instead of complete answerhood. It is more demanding in that it does not allow for redundancy in the answer. It requires that only information is provided which is exclusively directed towards the resolution of the issue raised by the question in the context. I will demonstrate that the new notion of contextual relatedness, together with our familiar notions of consistency and entailment, gives us the means to expand the fields of application of logical semantics to territories that used to be reigned by pragmatics.

Eva Hajicova (Prague)
How many levels of information structure are necessary?

A single level of the underlying structure (meaning) of the sentence is sufficient to account for all the linguistically patterned and semantically relevant phenomena connected with the topic-focus articulation (information structure) of the sentence, if the following requirements are met:

  1. linguistic meaning (accounted for by tectogrammatical representations, TRs) is distinguished from (extralinguistic, cognitive) content;
  2. the primitive notion of ‘contextual boundness’ (CB) is introduced in TRs, where CB nodes refer to well established entities "spoken about" and supposed by the speaker to be easily accessible to the hearer;
  3. underlying word order (the order of nodes in a TR) is distinguished from the surface one;
  4. a contrastive (part of) topic is distinguished.

Topic (T) and focus (F) can be then defined on the basis of (ii) and (iii): the verb itself and any of its immediate dependents belong to topic iff they are CB. Iff node n belongs to T, so do all nodes dependent on it (with a specific proviso for cases where all these items are CB). The rest of the TR is F.

Fritz Hamm (Tübingen) and Michiel van Lambalgen (Amsterdam)
Nominalization, the progressive and event calculus

In this paper we present a calculus of events based on recent work by M. Shanahan and show how to apply this calculus to the semantic analysis of two types of nominalizations and the progressive. M. Shanahan developed his calculus for completely independent reasons. He intended to give a logical account of the common sense informatic situation for a mobile robot. In contrast to recent work in event semantics however these notions are rigorously axiomatized and therefore result in a calculus of events.

Motivated by Vendler’s observations about two types of nominalizations in English we enrich Shanahan’s calculus by introducing conjunctions, disjunctions and negations of fluents. Again empirical considerations lead us to further subdivide the domain of events into the domains of event types and event tokens. We show that it is possible to introduce for each of these domains a notion of negation, one resulting in an antonym with respect to the set of event types, the other in a fluent. Furthermore we demonstrate that the calculus is consistent by constructing a model. The characteristic non-monotonic behavior of expressions in the progressive will be analyzed by using circumscription.

Kerstin Hansson (Lund)
Modifikation durch Adverbiale der Art und Weise aus semantischer und konzeptueller Sicht

Die nähere Untersuchung der präzisen Bedeutung der Adverbiale der Art und Weise ("A&W" abgekürzt) in verschiedenen Kontexten läßt eine erstaunliche Vielfalt an Lesarten zutage treten (1)-(5).

    1. Sie sammelte laut die Nägel auf. (Prozeßteil eines Ereignisses)
    2. Sie färbten den Stoff schön ein. (Resultatsteil eines Ereignisses)
    3. Sie färbten den Stoff großzügig ein. (Quantität eines inkorporierten Objekts)
    4. Das Auto fuhr schnell. (Geschwindigkeit, ein Parameter des Prozesses)
    5. Sie hatten das Stück gründlich geübt. (Iteration, ein Parameter der internen Struktur eines Ereignisses)

Die Beobachtung, daß sich die Variation weitgehend auf eine Korrelation zwischen sprachlichen Faktoren des Satzkontextes und den Lesarten der A&W zurückführen läßt, deutet aber auf eine prinzipienbasierte Modifikationsweise hin. Ausgehend von einer der gängigen Modifikationsprinzipien der Literatur, der Theta-Unifikation, wird eine Alternative zur theoretischen Erfassung der A&W-Modifikation vorgestellt, die die erforderliche Flexibilität gewährleistet, ohne dabei auf ein einheitliches Format verzichten zu müssen. Der Vorschlag beruht auf der Annahme einer grundsätzlichen semantischen Unterspezifikation. Die semantische Repräsentation wird auf eine konzeptuell begründete Darstellung abgebildet, in der auch kontextsensitive Bedeutungsvariationen erfaßt werden können. In diesem Zusammenhang werden die semantischen und syntaktischen Faktoren des Satzkontextes, die mit den Lesarten korrelieren, präzisiert. Anhand der Bedeutungsdifferenzierungen der A&W wird abschließend auf ihre Relevanz für Hypothesenbildungen zur Verbsemantik hingewiesen.

Klaus von Heusinger (Konstanz)
Fokus, Hintergrund und Unterspezifizierung

Semantische Theorien übersetzen fokussensitive Ausdrücke in Quantoren über Alternativen zu dem fokussierten Ausdruck. Diese Sicht führt jedoch zu Problemen, wenn anstelle von Eigennamen andere Ausdrücke fokussiert werden, wie in (1). Eine Quantifikation über Alternativen zu der Eigenschaft "amerikanisch" in (2) berücksichtigt zu wenige Alternativen, da sie aufgrund der Einzigkeitsbedingung der definiten NP alle Freundinnen ausschließt, die nicht einzig bezüglich einer Nationalität sind. Die intuitiv plausible Paraphrase (3) schließt hingegen alle Freundinnen in den Bereich ein, unabhängig davon, ob sie einzig bezüglich einer Nationalität sind oder nicht.

Ich schlage eine neue semantische Repräsentation der Informationsstruktur vor, in der der Satzinhalt in die beiden Informationseinheiten "Vordergrund" und "Hintergrund" übersetzt wird. In dieser Semantik wird nicht mehr über Alternativen zu dem Fokus quantifiziert, sondern fokussensitive Ausdrücke werden als Funktionen zwischen den beiden informationellen Einheiten dargestellt. Damit läßt sich die intutive Bedeutung (3) herleiten und es ergibt sich eine natürliche Einbettung von der Informationsstruktur in die Diskursstruktur.

(1) Luise hat nur der [AMERIKANISCHEN]F Freundin ein Foto von Peter gezeigt.
(2) Für keine Nationalität außer der amerikanischen gilt: Luise hat der einzigen Freundin mit dieser Nationalität ein Foto von Peter gezeigt.
(3) Für keine Freundin von Luise außer der amerikanischen Freundin gilt: Luise hat ihr ein Foto von Peter gezeigt

Wolfram Hinzen (New York)
DRT without modes or anchors

This talk sets out to criticize and ultimately dispense with two devices that have been essential in both Kamp’s (1990) and Asher’s (1993) modeling of attitudinal structure and content: external anchors, which were meant to account for the philosopher’s notion of wide content, and mode-indicators, which are particular conditions on discourse representation structures intended to mark semantically relevant differences between the attitudes. I argue that the notion of wide content is unneeded under a certain perspective on what discourse representation structures represent, and that mode-indicators presuppose the analysis of attitudinal differences they are asked to provide; it is argued that to obtain the latter not a modification of the object language is needed but of the semantics, which gives rise to see DRT more in line with semantic theories built on epistemic notions, like probabilistic semantics.

Martin Jansche (Columbus)
A uniform semantics for Mandarin Chinese wh-NPs

The question of what kinds of interpretations Mandarin Chinese wh-elements receive within a sentential context has been answered by the recent literature in roughly the following way: interrogative, existential, universal, and other ("donkey") bound uses of wh-elements are the result of binding by appropriate operators. However, none of these proposals provide a uniform semantics for the bound wh-elements, nor do they make it clear what binding of a wh-element amounts to. I address these issues by providing a formal account of Chinese interrogative NPs that gives them uniform lexical entries and accounts for their different uses and scopes in terms of certain closure operators associated with embedding elements. All lexical entries for NPs are assigned one polymorphic type and given a denotation that allows for the strictly compositional accumulation of the information contributed by wh-NPs. An exportation function makes sure-in connection with the lexical semantics of the embedding elements-that embedded wh-NPs receive appropriate scopes, allowing or blocking percolation of wh-information into a higher clause. It is important that wh-elements are not translated as existentially quantified (or otherwise closed off) variables, but rather as (partial, type-lifted) identity functions. This approach extends easily to account for existential (universal, etc.) uses of wh-NPs by postulating existential (universal, etc.) closure operators within the lexical meaning of embedding elements such as negation or (other) modals. This lexically uniform account of existential wh-elements avoids certain locality problems associated with the licensing conditions of alternative accounts.

Ingrid Kaufmann (Düsseldorf)
Argument-Linking bei resultativen V-V-Komposita im Chinesischen

Das Chinesische weist resultative V-V-Komposita auf, deren Argumentrealisierung der in anderen Sprachen vorliegenden Beschränkung widerspricht, daß Resultatsprädikate nur über das direkte Objekt prädizieren dürfen (1), (2b). Darüber hinaus finden sich transitive V-V-Komposita, die generelle Annahmen zum Argument-Linking verletzen: In kausativen Konstruktionen kann das direkte Objekt als Agens und das Subjekt als Patiens der von V1 kodierten Handlung interpretiert werden (2c), oder es kann ein Subjekt auftreten, das kein Argument von V1 oder V2 ist (3).

(1)
Youyou pao-lei-le
Y. run-tired-ASP

Youyou rannte (sich) müde.

(2)
Lisi zhui-lei-le Zhangsan.
L jag-müde-ASP Z

a. Lisi jagte Zhangsan und Zhangsan wurde müde.
b. Lisi jagte Zhangsan und wurde müde.
c. Zhangsan jagte Lisi und wurde müde (wegen Lisis Verhalten während der Jagd).

(3)
Zhangsan-de si ku-hong-le Lisi-de yanjing
Z’s death cry-red-ASP L’s eye

Lisi cried over Zhangsan’s death and as a result his eyes got red. (Gu 1992, p. 28)

Der zentrale Unterschied zwischen chinesischen V-V-Komposita und Resultativkonstruktionen im Englischen oder Deutschen besteht darin, daß die kausale Ereignisstruktur im Chinesischen nicht durch die Argumentstruktur reflektiert werden muß, da sie durch die morphologische Struktur des Verbs transparent ist. Deshalb kann bei koreferentem Agens und Patiens das Resultatsprädikat über das Subjekt prädizieren. Darüber hinaus müssen nicht alle semantischen Argumente ins Theta-Raster eingehen, solange die Argumente der Kompositumsteile durch semantische Koindizierung kontextuell identifizierbar sind. Ähnlich wie Li (1995) gehe ich davon aus, daß die Argumente von V-V-Komposita bei kausaler Ereignisstruktur "eventive Rollen" tragen, die durch die Teilereignisse lizensiert sind. Sofern die kontextuelle Identifizierung der Argumente der Kompositumsteile gewährleistet ist, darf ein Partizipant des verursachenden Teilereignis, der nicht Argument von V1 sein muß, als Subjekt mit der eventiven Rolle ‘Causer’ realisiert werden. Eine solche Trennung der eventiven und thematischen Rollenhierarchie ist in Sprachen, in denen die Ereignisstruktur über das Theta-Raster kodiert sein muß, ausgeschlossen.

Andrea Kowalski (Saarbrücken)
Zur Bedeutung (freier) Ausnahmephrasen im Deutschen

Es wird eine semantische Analyse vorgeschlagen, die dem Faktum gerecht wird, daß Ausnahmephrasen auch mit nichtuniversell quantifizierenden NPn ‚assoziieren‘ können. Ihre semantische Funktion besteht darin, die Gültigkeit der Quantifikation einzuschränken, wobei drei Bedeutungskomponenten zu erfassen sind: Beschränkung der Quantifikation, Ausnahmestatus und Domäneninklusion. Die von Ausnahmephrasen ausgeübte Einschränkungsoperation kann nun nicht wie in Moltmann (1993), (1995) an den in generalisierten Quantoren enthaltenen Mengen ansetzen, weil dies bei nichtuniversellen Quantoren (je nach Größe der Ausnahmemenge) zu schwache Wahrheitsbedingungen ergibt. Deshalb werden Ausnahmephrasen als Domänensubtraktoren analysiert, so daß die vom Determinator festgelegten Relationen nach der Ausnahmeoperation vorliegen müssen. Zusätzlich wird der Ausnahmestatus als direkter Bedeutungsbestandteil behandelt. Im Falle universeller Quantifikation folgt damit die Falschheit der unbeschränkten Quantifikation. Zugleich wird diese Analyse dem Vagheitsspielraum von Ausnahmestrukturen mit nichtuniverseller Quantifikation gerecht, indem sie den Wahrheitswert der unbeschränkten Quantifikation offen läßt bzw. ihn von der Größe der Ausnahmemenge abhängig macht. (Dies i.U. zu von Fintels (1993), (1994) Restriktivitätsbedingung, die auch bei kleinen Ausnahmemengen inadäquaterweise Falschheit der unbeschränkten Quantifikation fordert und bei mehrelementigen Ausnahmemengen nicht für alle Ausnahmen den Ausnahmestatus garantiert.) Demgegenüber wird Domäneninklusion als Bedingung für die Defininiertheit des Denotats von Ausnahmestrukturen aufgefaßt.

Philippe Kreutz (Bruxelles)
Perception and identity

The aim of my talk is to shed some light on perception report statements and more specifically on their semantic behavior with respect to the principle of substitution of extensionally equivalent expressions (PSE). With infinitival or -ing-form complements, predicates like to see, to hear or to feel obey (PSE), except with respect to propositions. But, in epistemic that-reports, these very same predicates do not satisfy (PSE). This general contrast between perceptual and epistemic attitude reports cannot be fully grasped in terms of the de re/de dicto distinction. First, the distinction itself is problematic for epistemic reports. This means that, from a pragmatic point of view, de re attributions don’t always induce referential transparency. Secondly, even if epistemic and perception reports were alike under their de re interpretation, we would still have to explain why perception reports have no de dicto reading. We will rather argue that the behavior of perception reports towards (PSE) lies both in the occurrential nature of perceptual modalities and in the analogical representations these modalities induce. We rely on the fact that identity questions simply do not arise in analogical representations because these are basically about occurrences and not about types of occurrences. As Stenning and Inder have previously argued, even when types are allowed in an analogical representation (via categorization), the question of their referential identity is solved by the representation itself. We thus promote a representational and cognitive approach on a semantic issue. Even a psychological understanding of Fregeanism (relying on a distinction between direct and indirect reference) could not explain the asymmetry between epistemic and perceptual contexts. Likewise, neither situation semantics, nor phenomenology (assuming a non-representational dimension of perception) make a sufficient contribution to this problem.

Annette Leßmöllmann (Hamburg)
Why there are no round ways: shape adjectives and shape representation

This paper deals with the question how German adjectives that denote shape properties get their meanings. The discussion focuses on clashes like in (1):

(1) *der runde Weg (*the round way)

(1) may neither assign a shape property nor the property of being circular, as in the compound Rundweg (circular route).

Considering psychological theories on object recognition, conceptual principles are formulated which constrain the applicability of shape terms. It is shown that clashes may be predicted on the basis of generalizations over geometric properties of concrete objects. The adjective rund (‘round’) in predicative and attributive use only denotes shape properties of contours of objects, but not properties of axes. Hence, round differs from adjectives like gebogen (‘bent’). On the other hand, it falls into the same class like eckig (‘angular’), which also may not be used for the shape of the elongation axis:

(2) *der eckige Deich (*the angular dyke)

These clashes may be predicted starting out from general hypotheses on object representations.

It is also shown that the examples in (3) do not justify the assumption of a polysemous lexical entry for rund (i.e. a ball-like versus a cylinder-like roundness):

(3)
a. runder Turm (round tower)
b. runder Luftballon (round balloon)

Taking into account the interaction between linguistic and spatial format, the apparent variability of rund may be reduced to one semantic representation.

Sebastian Löbner (Düsseldorf)
Van der Auwera's puzzle: finally already solved?

In his 1993 paper 'Beyond Duality', van der Auwera presents the observation that 'already' and 'finally' are incompatible in sentences such as 'Peter is already/finally in Madrid'. "Finally", he argues "expresses that the change into the positive state has happened relatively late. 'Already' expresses that the change into the positive state has happened relatively early or better, [...], it excludes that the change has happened relatively late." This observation, van der Auwera argues, invalidates the duality analysis proposed in Löbner (1989), according to which 'already(p)' is equivalent to 'no longer(not-p)': 'finally(p)' entails 'no longer(not-p)', this in turn would entail 'already(p)', thus leading to a contradiction in van der Auwera's eyes.

I will argue that 'finally' expresses a "relatively" late change in a different sense than 'already' expresses a "relatively" early change. 'Finally' has an expressive, i.e. non-propositional, meaning component to the extent that the speaker expresses a subjec-tive negative attitude to the amount of time it took for p to come about. The propositional meaning is analyzed as expressing a phase transition from not-p to p (same as 'already') with the additional condition that the initial phase of not-p started long before the given time of reference. 'Already', in contrast, has no corresponding expressive meaning component. Its sense of relative earliness derives naturally from its propositional meaning if the phase quantifier analysis is adopted for the particle. It follows from the analysis that 'finally(p)' indeed entails 'already(p)', since entailment is only a matter of propositional meaning.

The apparent conflict between the two particles is one in terms of foregrounding. Both particles foreground one of the two phases of the not-p/p phase transition scheme. 'Already(p)' expresses the state of affairs in terms of p, hence foregrounding the second phase of the not-p/p scheme. 'Finally(p)', due to its prominent expressive meaning component, emphasizes the first phase, not-p.

References:
Löbner, Sebastian: 1989, 'German schon - erst - noch: an integrated analysis', Linguistics and Philosophy 12: 167-212.
van der Auwera, Johan: 1993, ' 'Already' and 'still': Beyond duality', Linguistics and Philosophy 16: 613-653.

Cécile Meier (Tübingen)
Konsekutive Komparationskonstruktionen

In meinem Vortrag werde ich eine Analyse für Konstruktionen mit so ..., daß und mit zu ..., als daß im Rahmen der Theorie von Stechow (1984b) vorschlagen. Ich werde dafür argumentieren, daß sowohl Konstruktionen mit so ..., daß wie auch mit zu ..., als daß Vergleiche von zwei Ausmaßen ausdrücken. Das erste Ausmaß ist durch den übergeordneten Satz determiniert. Das zweite Ausmaß ist das minimale (für so) oder das maximale (für zu) Ausmaß aus einer Menge von Ausmaßen, die durch ein "verstecktes" Konditional determiniert sind. Der untergeordnete Satz stellt generell das Konsequens dieses Konditionals, der übergeordnete Satz das Antecedens. Die Vergleichsrelation ist die "größer(-gleich)"-Relation. Diese Idee exemplifizieren (1) und (2).

(1)
a. Bertha ist so groß, daß sie über die Mauer schauen kann.
b. "Das (max.) Ausmaß e, für das gilt: Bertha ist e-groß ist größer oder gleich groß wie das MINIMALE Ausmaß aller Ausmaße e für die gilt: Wenn Bertha e-groß ist, kann sie über die Mauer schauen."

(2)
a. Das Packet ist zu schwer, als daß man es hochheben könnte.
b. "Das (max.) Ausmaß e, für das gilt: Das Packet ist e-schwer ist (echt) größer als das MAXIMALE Ausmaß aller Ausmaße e für die gilt: Wenn das Packet e-schwer wäre, könnte man es hochheben."

Den Vorschlag gilt es formal für Konstruktionen mit negativen und positiven Adjektive zu präzisieren und an relevanten Beispielen zu überprüfen.

von Stechow (1984b): My reaction to Cresswell’s, Hellan’s, Hoeksema’s and Seuren’s comments. Journal of Semantics 3, pp. 183-199.

Ralf Naumann & Anja Latrouite (Düsseldorf)
A modal logic for the interpretation of aspectual phenomena in natural language

An assumption going back to Vendler (1967) that underlies many aspectual theories is that a verb or a VP can be modified either with in- or with for-adverbials. Yet this assumption proves to be wrong. First, there are verbs like wash that admit of both types of adverbials (wash a shirt in/for ten minutes). Second, in Japanese there are verbs like hue (‘increase’) which can neither be modified with in- nor with for-adverbials, in contrast to their translations in English or German. This raises the further problem of how this difference in aspectual behavior between a verb and its translation in other languages can be explained. Moreover, in many languages the (aspectual) interpretation of verb depends on the particular (morphological) form that is used. For instance, there is strong evidence that in Tagalog the interpretation of a verb depends on the voice-form used. The object-voice of accomplishment-verbs like read or eat only has what is usually called a single event reading (that corresponds to the usual interpretation of its English translation). The agent-voice, on the other hand, has at least two interpretations. The first corresponds to a form of iterative reading and the second is similar to a partitive reading.

(1)
Nagbasa ako ng libro
Realis-AV-read Nom:I Gen book
I read in a book.

(2)
B-in-asa ko ang libro
B-Realis/OV-read Gen:I Nom book
I read the book (completely).

In this talk we will present a theory in which a more fine-grained analysis of verbs is possible that admits an explanation of the above data. The analysis is based on the assumption that non-stative verbs express changes. The notion of a change comprises at least two complementary aspects. The first one views changes as transformations of states: a state s is transformed into a state s'. The disadvantage of this perspective, underlying Dynamic Logic (DL), is that changes are not objects of some domain but are interpreted as binary relations on the domain S of states. The second perspective, adopted in event-semantics, views changes as objects, namely as elements of a separate domain E of events. What is missing is the procedural perspective underlying DL. In our approach both perspectives are combined. Besides a transition-structure S, there is an eventuality-structure E with an underlying domain E of events modeling the perspective of changes as objects. A modal language L is defined that combines the dynamic-temporal logic of Van Benthem et al. (1993) with the modal interval logic of Halpern & Shoham (1986).

Aspectual differences are explained in terms of the properties the execution-sequences of events have. Different aspectual classes correspond to different types of programs that are distinguished by the properties they have. Imperfectivity (durativity) is explained as a form of invariance-property whereas perfectivity (terminativity) is explained in terms of non-invariance properties of execution-sequences. Aspectual differences between a verb and its translation in another language (hue vs. increase) are explained by assuming that although the verbs can be used to describe the same type of complex event in the world, the verbs can denote different parts of events of this type. At the formal level these differences are captured by applying various closure operations to a given program that corresponds to the complex event at the level of a change as a transformation of state. The same explanatory mechanism is used for the difference in interpretation between the object-voice and the agent-voice forms in Tagalog.

Ralf Nüse (Nijmegen)
Allgemeine Bedeutungen für an auf und unter

Nach allgemeiner Auffassung gibt die Bedeutung einer Präposition eine jeweils spezifische Region des Relatumobjekts an, in der sich das Thema befindet. Die entsprechenden Referenzregionen von auf und unter werden dabei mit Bezug auf die Vertikale, die von an dagegen mit Hilfe topologischer Begriffe definiert. Derartige Bedeutungen decken allerdings nicht alle Verwendungen dieser Präpositionen ab, und sie sind auch nicht für die jeweilige Präposition spezifisch. Zudem sind sie prinzipiell nicht in der Lage, die Trennung und die Überlappung der Referenzregionen von an, auf und unter bei verschiedenen Arten von Relatumobjekten zu rekonstruieren (z.B. überlappen an und unter bei Plakaten, aber nicht bei Flaschen und Tafeln, dafür überlappen an und auf bei Flaschen und Tafeln, aber nur bedingt bei Plakaten). Diese Probleme können gelöst werden, wenn man die Bedeutung bzw. Referenzregionen dieser Präpositionen auf der Basis von Alltagskonzepten wie OBERFLÄCHE, SEITE und UNTERSEITE definiert. Wie sich empirisch nachweisen läßt, ist die Oberfläche eines Objekts im Alltag so etwas die Oberseite und/oder die größte Fläche des Objekts, die keine Unter- oder Rückseite ist, eine Seite ist orthogonal zur Unterseite, und die Unterseite ist diejenige Fläche, die "nach unten weist" und/oder als Ganzes in Kontakt mit einer anderen Fläche ist. Die so definierten Objektregionen korrelieren genau mit den entsprechenden Referenzregionen von auf, an und unter. Zudem lassen sich damit auch einige subtile(re) Interpretations- und Gebrauchspräferenzen dieser Präpositionen erklären.

Irene Rapp (Tübingen)
Semantische Struktur und Argumentrealisierung

Thema des Vortrags ist das Linking von Verbsemantik und Argumentrealisierung. Meine Verbklassifikation gibt thematische Struktur, Ereignisstruktur und Kausalitätsstruktur durch eine einzige semantische Repräsentation wieder. Die Verben sind in eine sublexikalische Prädikat-Argument-Struktur dekomponiert. Basisprädikate (mit Individualargumenten) geben die thematische Struktur wieder:

(1)
streicheln: DO (x,y) x = Actor, y = Patiens
wissen: PSYCH (x,y) x = Experiencer, y = Thema

Die komplexen Prädikate nehmen andere Prädikate als Argumente. Aus ihrem Vorhandensein (bzw. Fehlen) lassen sich Ereignisstruktur und Kausalitätsstruktur ableiten:

(2)
erfahren: BECOME (PSYCH (x,y))

(3)
mitteilen: CAUSE (DO (x,y), BECOME (PSYCH (z,y)))

Meine These ist, daß die Ereignisstruktur keine Bedeutung für das Linking hat, wohl aber die thematische Struktur und in bestimmten Fällen die Kausalitätsstruktur. Das Zustandswechselverb erfahren realisiert seine Argumente ganz parallel zum Zustandsverb wissen, da beide das gleiche Basisprädikat besitzen. Konflikte können sich ergeben, sobald mehrere Basisprädikate vorhanden sind; also bei kausativen Verben. So müßten der thematischen Struktur zufolge bei mitteilen sowohl der Actor als auch der Experiencer extern realisiert werden. Hier zeige ich, daß das erste unter CAUSE eingebettete Prädikat bei der Argumentrealisierung prinzipiell Vorrang hat. Für die Argumente des zweiten Prädikats werden Ausweichposition (hier in Form des strukturellen Dativs) geschaffen.

Marie-Eve Ritz (Nedlands)
Focus on the English present perfect

This paper proposes a unified account of the English present perfect (PP), a long standing problem in semantic analysis, and discusses how, taking into account recent treatments of such phenomena as focus and presupposition, the semantics of the tense can help explain some pragmatic effects.

The paper begins by considering some problematic examples and discusses briefly the variety of approaches that have been taken to explain the meaning(s) of the tense. It then argues that a unified semantic analysis is possible, and the account proposed builds on Moens’ (1986), and Vet’s (1981) aspectual representations. The central thesis defended here is that PP sentences can be represented as denoting a consequent state which follows a culmination point, and that the culmination itself always represents a change between two opposite states. Thus the negation of the stated proposition is always presupposed as having been the case before the culmination took place. In other words, a PP sentence gives a hearer the instruction to find a culmination point which represents a change from a negative state to its positive counterpart. The paper then proceeds to examine ‚always‘ sentences in the PP as well as negated PP sentences, and builds in the above representation for these types of sentences using DRT notation and Van der Sandt’s (1988), (1992) approach to presupposition. Finally, some pragmatic extensions are considered. It is argued that when we take into account such phenomena as topic and focus, the proposed analysis explains the behavior of the PP, thus providing a basis for understanding constraints imposed by the position of adverbials in the sentence as well as manipulation of semantic features for effect by speakers.

Robert van Rooy (Amsterdam)
Presupposition satisfaction

The satisfaction approach of Karttunen and Stalnaker towards presuppositions goes arguably a long way in giving an explanatory solution to the projection problem for presuppositions, the problem to determine the presuppositions of complex sentences given the presuppositions of their parts. The idea is (i) that every sentence should be interpreted with respect to a context that represents what is commonly assumed, (ii) that a sentence can only be appropriately asserted in this context, if the triggered presupposition is already "part of", or satisfied by, the context, and (iii) that the presupposition induced by complex sentences should be explained by the way the context changes during the interpretation of the sentence. It is generally agreed that the satisfaction approach towards presupposition is intuitively plausible, but also that the usual way the approach is implemented (Heim, Van Eijck, Beaver, ...) gives rise to some implausible predictions. Van der Sandt and Geurts concluded that the approach is wrong and has to be replaced by an alternative account, making heavy use of accommodation and a more structured representation of context. Instead of arguing for the satisfaction approach by defending some implausible predictions, or developing an empirically more adequate alternative based on conceptually shaky grounds, I rather want to argue that the satisfaction approach itself need not give rise to such implausible predictions. My argument will be based on the assumption that (i) contexts should indeed contain more information than they are usually said to contain (modal subordination), or are more structured than they are usually assumed to be (non-local satisfaction), and (ii) that the idea that contexts represent what the speaker is assuming for the sake of conversation should be taken more seriously than is usually done (presupposition as a propositional attitude).

Antje Roßdeutscher (Stuttgart)
When-sentences-temporal anaphora and non-accidental dependency

I will review some promising approaches to sentences of the form "When A happened, B happened". These are nowadays successfully accounted for in terms of "discourse relations". I will present some still puzzling examples and counterexamples that will serve to guide the presentation of the new account during the talk.

Justification of when in context.

In order to interpret a particular occurrence of when, its use must be justified both with respect to the Common Ground and with respect to the verbal context, the verbal descriptions of A and B. Context in the sense of shared knowledge and information structure is decisive for whether when has a temporally localizing reading only (see question-answer-pair (1))-or contributes a relation of non-accidental dependency of A and B in addition to that (see (2)). Non-accidental dependency mirrors counterfactual reasoning, (see (2')). (The interfering readings of "narrative" when-sentences can be accounted for along these two dimensions, too.)

(1)
When did B happen?-B happened, when A happened.

(2)
What happened, when A happened?-When A happened, B happened.

(2')
If A hadn’t happened, B would not have happened.

In the case of (2) the lexical details of the verbal descriptions become crucial for justifying the particular type of non-accidental dependency. These types reconstruct discourse relations such as ‘response’, ‘elaboration’, ‘enablement’. Besides tense, aspect and Aktionsart, intentionality is a decisive dimension.

Non-accidental dependency also provides a clue to understanding the lexical field of temporal conjunctions bevor, während, als, and nachdem.

Uli Sauerland (Chiba)
A uniform approach to DP quantification

Current theories of DP quantification posit a dichotomy between interrogative quantifiers, which are allowed to quantify over choice or Skolem functions (e.g. Engdahl (1980), (1986), Chierchia (1992)), and non-interrogative quantifiers, which are restricted to quantification over individuals. Evidence from the distribution of Condition C effects (Chomsky (1993), Fox (1994)) and identity of trace effects in Antecedent Contained Deletion (Sauerland (1998)) shows that the dichotomy is not only conceptually, but also empirically problematic. This talk develops a uniform semantics for DP quantification, in which all quantifiers uniformly quantify over choice functions. Von Stechow (1996)’s earlier proposal along these lines is shown to make wrong predictions for proportional and cardinal quantifiers, as well as cases involving a bound variable. As a refinement of von Stechow’s proposal, I propose to restrict quantification over pointwise different choice functions.

Philippe Schlenker (Cambridge)
Propositional attitudes and embedded indexicals

The value of a clause embedded under a propositional attitude verb is standardly taken to be a proposition (=a set of possible worlds) or, in recent accounts, a property (=a set of possible individuals). We argue that this is incorrect, and that the object of an attitude is (irreducibly) a Kaplanian character, i.e. a function from contexts to propositions. If K is a character, Peter believes that K has the following truth-conditions:

(1)
[[Peter believes that K]] is true at w* iff for all Peter knows at w* he is in a context (of speech / of thought) c and in a world w such that K(c)(w).

1. Our first argument is that there are across languages indexicals which are evaluated with respect to the context of the reported speech act rather than of the matrix clause. Thus in Aghem what is literally the woman said to him that she much like YOU may mean the woman said to him that she liked HIM a lot, where you in the embedded clause refers to the hearer of the reported speech act.

2. The semantics in (1) gives a straightforward account of de se readings: in Heimson believes he is Hume, he is interpreted like a 1st person pronoun which is evaluated with respect to the context of the reported thought act:

(2)
[[Heimson believes he is Hume]] is true at w* iff for all Heimson knows at w* he is in a context of thought c and a world w such that author(c)=Hume at w.

3. Finally, we account for the fact that he rather than I appears in (1) by showing that the phenomenon of ‘sequence-of-tense’ has a counterpart in the pronominal domain: ‘sequence-of-person’ in (2) which shifts morphologically a 1st person to a 3rd person, without changing its semantic interpretation.

Kerstin Schwabe (Berlin)
About indefinite expressions in coordinative structures

The paper aims to show that shared indefinite expressions in coordinative constructions may differ with respect to their referential properties, that this is due to their different information structural status and that the latter determines the syntactic representation of such coordinations as elliptical or ATB constructions.

Following the Theory of Focus Interpretation of Rooth (1992), indefinites in shared focused constituents must be beyond the actual coordination structure-cf. (1), whereas indefinite expressions not being in focus may remain in coordination.

(1)
[CP1c[CP1bHans hat Anna und Fritz hat Paula] [einen Schüler
vorgestellt
]F1]
Hans has to Anna and Fritz has to Paula a pupil introduced.

Er soll ein guter Pianist sein.
He is considered to be a good pianist.

If the shared indefinite expression is in focus, it may refer uniquely. This means that there are not any other entities denoted, which have the same predication. The unique reading presupposes that the subjects on the one hand and the objects on the other hand in CP1b form a set. This set formation is only possible if the actual coordination is not in focus. Additionally, the indefinite may get a distributive reading. This is gained by the application of a distributive operator to the respective set.

Additionally to focus theoretically based arguments, the paper will discuss interesting pieces of evidence that support the claim that shared focused constituents have to be beyond the coordination phrase. They concern in addition to the possible unique reading of shared focus indefinites, the reference of pronouns if they are in the presentational focus domain and number agreement.

Pieter A. M. Seuren (Nijmegen)
Intensional committal discourse semantics: a five-finger exercise

This presentation is part of an overall research program aiming at a radical re-orientation of natural language semantics in terms of socially binding commitments (speech acts) and a cognitive information update system. Compared with model-theoretic semantics, the role of speech acts is greatly enhanced, at the expense of logic. Contextually bound update (incrementation), with open access to available knowledge and subject to the peculiarities of human cognition, is an essential component. Formalization is much less a question of logical or model-theoretical calculus than of implementation. Inevitably, the role and position of pragmatics will have to be redefined, rather to its disadvantage. Time limits prevent an all-round survey of all the issues and arguments involved. Rather, I intend to focus on a few specific instances, the semantics of or and if, of some and of reflexivity. For or and if an important problem is the fact that:

(1)
a. Either John has no car, or he does have one but then it’s in the
workshop
b. If John has a car, it’s in the workshop

are standardly considered true if, besides one or more cars in the workshop, John also has one parked up the road. But in language that makes them false: the car parked up the road is a ‘gatecrash falsifier’. The solution proposed rests on the notion of commitment. As regards (1a), in a disjunction ‘A or B’ the speaker commits himself to a choice between A and B as alternative incrementations. If B is ‘John has a car and it’s in the workshop’, then, once A (‘John has no car’) has been rejected, B MUST be added and anything entailing that John has a car that is not in the workshop is excluded. Likewise for (1b): ‘if A then B’ commits the speaker to incrementing B in case A is incremented. As regards some, an explanation is provided for the well-known implicature not all resulting from the use of some of the in terms of the fact that human cognition equates ‘subset’ with ‘proper subset’, as part of a more general ‘Principle of Non-Reflexivity’ which avoids reflexivity and requires distinctness of terms in n-termed relations. This explains why ‘Some of the children laughed’ is consistent with ‘All children laughed’ (though it suggests the opposite), while ‘Only some of the children laughed’ is not. The Principle of Non-Reflexivity explains why ‘Clare envies all the girls in her class’ does not entail that Clare envies herself (which eliminates the barber paradox).

Benjamin Shaer (Montreal)
"Sequence of tenses" and sequences of tenses

In this paper, I shall be attempting to characterize the nature of the temporal dependencies between the clauses of ‘past under past’ (henceforth Pa\Pa) sentences like those in (1):

(1)
a. Joe said that he finished his work.
b. Joe said that he enjoyed the party.

I shall do so by making two basic claims about them. The first is that the range of temporal dependencies displayed in these sentences is far broader than commonly assumed, having too wide a distribution (i) to be plausibly determined by some ‘SOT rule’ (e.g. Ogihara 1996); or (ii) to be seen as operating ‘only ... in intensional contexts’ (Abusch 1997, p. 29) or as subject to an ‘upper limit constraint’ (ULC) on the denotation of tenses (ibid., p. 25). The second is that intrasentential temporal relations in Pa\Pa sentences can be explained in largely the same terms as those between sequences of past-tensed sentences (henceforth Pa+Pa sentences) like those in (2):

(2)
a. Jane climbed Mont Blanc. She walked up to the base camp.
b. Jane drank some water. She ate some chips. (ter Meulen 1995, p. 10)

The strategy here will be to show that the syntactic and semantic properties of tenses in Pa\Pa and Pa+Pa sentences have enough in common that it becomes plausible to attribute the differences in the interpretation of tenses in these two cases directly to differences between sequences of sentences and complement-clause sentences - in particular, to the differences in the semantic types that may respectively be associated with these structures. Accordingly, Pa\Pa sentences can be treated as a special case of Pa + Pa sequences.

Abusch, D. (1997). Sequence of tense and temporal de re. L&P 20, pp. 1-50.
Ogihara, T. (1996). Tense, Attitudes, and Scope. Dordrecht/Boston/London: Kluwer.
ter Meulen, A. (1995). Representing Time in Natural Language. Cambridge, MA: MIT Press.

Andreas Späth (Leipzig)
Determination und Semantische Form am Beispiel slawischer und deutscher DPs

Anhand des Russischen u.a. slawischer Sprachen sowie des Deutschen - welche sich hinsichtlich der Determination unterscheiden, wird gezeigt, welchen Beitrag DPs am kompositional-semantischen Aufbau der Satzbedeutung haben. Es wird die SF von Sätzen mit strukturell zweistelligen Verbprädikaten aus der LF-Konstellation im Rahmen einer Minimalistischen Syntax (Chomsky 1992) abgeleitet. Ausgehend von der Annahme, daß das Topik eines Satzes mit Definitheit assoziiert ist und damit die hierarchisch höchste Prädikation im Satz darstellt und die natürlich fokussierte DP im Skopus des Kommentar liegt, ist z.B. für den Satz Der Junge kaufte ein Buch bzw. Mal'tschik kupil knigu (russ.) die folgende semantische Repräsentation kompositional herzuleiten:

(1) $! x[[boy(x)]Ù [$ e[$ y[[book(y)]Ù [e inst[buy[x,y]]]]]]]

Dabei wird gezeigt, daß die Termanalyse für DPs es nicht ermöglicht, diejenigen Identifizierungs- und Kontextbedingungen zu repräsentieren, welche durch grammatische Mittel der Diskurseinbettung vorgegeben werden. Plädiert wird für die Quantorenanalyse von DPs. Es wird gezeigt, wie es im Prozess der semantischen Amalgamierung zur Skopusnahme über Argumentdenotate oberhalb der Proposition kommt und welche semantische Interpretation skopusbildende Spuren erhalten müssen. Die hier vorgeschlagene Analyse zum kompositional-semantischen Aufbau einer kontextinvarianten Satzbedeutung unterscheidet sich von bisherigen Analysen darin, daß der isomorphen Abbildung syntaktischer Strukturen auf die Semantik strikt der Minimalistischen Syntax folgt, wo berücksichtigt wird, daß lexikalische Einheiten über morphosyntaktische und semantische Informationen verfügen, welche an den Argumentstellen adressiert sind und in die entsprechenden Derivationspositionen projiziert werden. Somit kann in rein morphosyntaktisch und rein semantisch motivierte Bewegungen unterschieden werden, wobei letztere der Skopusnahme und damit der Subordination von Prädikationen im Satz dienen.

Heike Tappe (Hamburg)
'Die Linie trifft auf die Ostseite der Straße': "Mixed conceptual structures" in language production

The transition from concepts to lexical items is an essential component of language production. Research aiming at an understanding of the processes underlying this transition is facing three major problems: 1) the specification of nature of conceptual structures 2) the exploration of the structure of lexical knowledge, and 3) the gaining of insights into the interaction between conceptual and lexical structures, cf. Bierwisch & Schreuder (1992).

The basis for our investigations into these complex problems are spoken texts which are elicited in standardized settings. Verbalizers describe either the continuous, pixel-per-pixel genesis of sketch-maps (dynamic sketch-maps) or discrete sketch-maps (static sketch maps). Both kinds of stimulus material are presented in two conditions: In an on-line–i.e. simultaneous–condition the verbalizations take place during the respective presentation. In an off-line-posterior-condition, the verbalizations take place after the respective presentation. While describing sketch-maps, speakers can describe the representing entities, namely the graphical entities, or, rather, the represented entities, i.e. the entities intended by the creator of the sketch-map. In the data we find both: Firstly, utterances, in which graphical objects and events are referred to (‘now a rectangle is drawn, this form is hatched’). And, secondly, utterances, which describe real-world objects and events (‘but you drive straight on, now there is a crossroads, well, we have a street that goes from the West to the East’). Most interesting in the current context is a third type of utterances, in which both representational layers are accessed by the conceptualization component. These mixed utterances are subdivided into different groups. In most of them we find a predicate belonging to the graphical layer, e.g. "draw" and an argument belonging to the intended layer, e.g. "house". These cases of reference transfer are described according to the theory of enriched composition, following Jackendoff (1997). More complicated cases like, for example ‘where the line meets the big side of the street, there another line goes to the right, then over the street’ are only present in descriptions of dynamic sketch maps. They are analyzed as cases of procedural type coercion brought about by the simultaneous activation of both representational layers.

Bierwisch, M. & R. Schreuder (1992). From concepts to lexical items. Cognition 4, pp. 23-60.
Jackendoff, R. (1997). The Architecture of the Language Faculty. Cambridge, Mass: MIT Press.

Carla Umbach (Berlin/Hildesheim)
Zur Bedeutung von aber: Worin besteht der Kontrast?

Der Konnektor aber gilt als der allgemeinste Ausdruck zur Herstellung eines Kontrasts zwischen zwei Propositionen, über die Natur der Kontrastbeziehung selbst herrscht jedoch wenig Klarheit. Meistens wird angenommen, daß aus den Konnekten ein Widerspruch zu Normalitätsannahmen, Erwartungen o.ä. ableitbar sei. Dann wäre allerdings die Bedeutung von aber grundsätzlich nicht monoton beschreibbar.

Da aber offensichtlich mit der Topik-Fokus-Verteilung in den Konnekten interagiert, liegt es auf der Hand, die Ergebnisse zur Fokussemantik heranzuziehen. Tatsächlich läßt sich auf dieser Basis eine einheitliche Deutung für aber-Konstruktionen entwickeln, die den durch aber induzierten Kontrast klassisch monoton erklärt und die Verbindung zum Begriff des Kontrastfokus herstellt. Kurz: In dem mit aber angeschlossenen Konnekt wird eine (andere) Alternative der durch den Fokus des ersten Konnekts aufgespannten Alternativenmenge benannt und dabei implizit oder explizit für das Topik des ersten Konnekts ausgeschlossen.

Diese Deutung braucht auf keinerlei Widerspruch zu rekurrieren. Verletzte Normalitätsannahmen u. ä. lassen sich als für negative Aussagen typische Implikaturen erklären. Allerdings findet sich auch bei dieser Deutung ein Widerspruch, und zwar in den durch die aber-Konstruktionen evozierten Quaestiones: Es handelt sich immer um Entscheidungsfragen, die durch die Äußerung mit "ja und nein" beantwortet werden. Das weist deutlich daraufhin, daß Kontrast in erster Linie ein Phänomen der Textorganisation ist.

Helmut Weiß (Regensburg/Jena)
Die doppelte Negation: Einblicke in die Syntax-Semantik-Schnittstelle

Die doppelte Negation mit harmonischer Lesart, bei der zwei oder mehrere Negationsträger in einem Satz dennoch nur eine Negation ergeben, ist ein in sehr vielen natürlichen Sprachen vorkommendes Phänomen. Der Normalfall ist dabei, daß im Skopus der Satznegation bestimmte Indefinita als sog. inhärent negierte Quantoren (INQ) – z.B. niemand, nichts, kein – realisiert werden müssen. Im Vortrag wird eine – basierend auf der Kamp-Heimschen DRT, Diesings Mapping-Hypothese sowie dem Neg-Kriterium – Theorie der doppelten Negation im Bairischen vorgestellt, aus der sich eine einheitliche und präzise Erklärung für Form und Funktion dieser Konstruktion ableiten läßt. Die Theorie erlaubt einerseits die Erklärung universeller Phänomene sowie andererseits die Formulierung verschiedener Hypothesen für das Bairische, die als empirische Tests gelten können.

Damit wird z. B. die Erklärung einer universellen Lücke im Lexikon möglich: In vielen Sprachen ist die Negierung von Indefinita wie jemand etc. lexikalisiert (s.o.), es ist aber keine Sprache bekannt, die eine Form wie nalle als inhärente Negation von alle lexikalisiert hätte. Aus der hier vorgestellten Analyse der doppelten Negation folgt eine natürliche Erklärung für diese bislang unverstandene Restriktion.

Für das Bairische läßt sich mit dieser Theorie auch erklären, daß und warum es semantische Unterschiede zwischen Sätzen mit doppelter und einfacher Negation gibt. Außerdem erlaubt sie die Erklärung einer Vielzahl von Erscheinungen der bairischen Negationssyntax, die mit einer rein syntaktischen Analyse nicht zu leisten wäre.

Adam Z. Wyner (Ramat Gan)
Against the neo-Davidsonian event argument: A neo-Kimian alternative

We argue against the Neo-Davidsonian Event Theoretic analysis of adverbial modification (Davidson 1967 and Parsons 1990), where we predicate of an event variable and bind it existentially. (1a) is represented as (1b).

(1a) Brutus stabbed Caesar violently with a knife.
(1b) Ee [STAB(e) & AG(e)=brutus & TH(e)=caesar & VIOLENT(e) & WITH(e, knife)]

We claim that the event argument permits an illicit entailment in (2).

(2)
Every event in which Bill walked with Jill, Bill talked with Jill;
and every event in which Bill talked with Jill, Bill walked with Jill.
Every event in which Bill walked with Jill, Bill walked with Jill quickly.
Therefore, Bill talked with Jill quickly.

Second, we argue that the event argument is redundant. And third, the event argument does not behave anaphorically as other individuals; in particular, we can not refer to the very same entity, "Jill's kissing Bill on January 1, 1990", in modal contexts as in (3b).

(3a) Jill kissed Bill on January 1, 1990.
(3b) It could have happened with the participation of Will instead of Jill.

We introduce the Neo-Kimian approach in which there is no event argument, but rather a set of properties, relations, and individuals; the theory does not have the problems of Event Theory, nor those of Kim's (1976) original theory (see Bennett (1988) and Parsons (1990)). For discourse anaphora, we suppose, following Asher (1993), that we have a function which maps sets to abstract objects which serve as discourse antecedents. The analysis is an improvement over Event Theory, and it has broad ramifications in a variety of areas of current research.

Thomas Ede Zimmerman (Stuttgart)
Disjunction and free choice

This paper offers a solution to Hans Kamp’s (1973) problem of free choice permission: how can it be that sentences of the form "X may A or B" are usually understood as implying "X may A and X may B"? Unlike other approaches that locate the problem in the semantics/pragmatics interface, the solution to be presented is purely semantic, crucially depending on a modal account of disjunctions as lists of epistemic possibilities.

SINN-Startseite SINN-home