Contrastive Topic (CT), marked by prosody as in English and German (Topic-Focus contour) or by morphology as in Japanese (-wa) and Korean (-nun), is often accompanied by
implicatures.
Buering (1997), one of the most influential works on CT, employs the notion of Topic-value, which is a complex structure of alternatives, a set of sets of propositions. Buering claims that the CT-marked sentence is infelicitous if there is no unanswered question in its Topic value. This approach makes the wrong predictions for the so-called last answer problem, where all the presupposed questions are resolved.
Hara (2005) gets around the last answer problem by entertaining the idea that the CT-marking indicates the speaker's limited competence in terms of quantity implicature computation, namely by removing the opinionatedness assumption. However, this approach makes the wrong predition for a sentence involves a Topic-Focus structure.
We will offer a new formalization of Contrastive Topics. That is, CT presupposes a question, forms a simple set of Topic-alternatives, and gives rise to the conventional implicature that one of the Topic-alternatives is not known to be true. Our formalization does not run into the last answer problem and does not remove the speaker's competence assumption. Furthermore, we show that a simpler set of Topic-alternatives (a set of propositions rather than a set of sets of propositions) is sufficient to account for the distribution of CT.
I will present an account of Condition B effects, Strong Crossover effects, and certain puzzling restrictions on the interpretation of pronouns in VP-ellipsis.
The use of rhetorical questions as emphatic statements is attributed to their
literal meaning as a question. The proposed account of rhetorical questions
focusses on negative polarity items (NPIs), a characteristic of these
questions.
The integration of an NPI into a question greatly affects the set of
exhaustive answers to this question (i.e., the meaning of the question). For
yes/no-questions this introduces the presupposition that the corresponding
question without the NPI is already settled in the negative, which is seen as
the main impact of the NPI and the reason for the rhetoricity of the question
(Krifka 1995, Rooy 2003).
It will first be shown that for wh-questions, however, the integration of an
NPI does not settle the corresponding question without NPI in the same way.
It is argued that rhetoricity already emerges from the general
threshold-lowering effect of NPIs, which makes in particular wh-questions too
general to be of interest to the speaker (in a literal interpretation).
Second, I will then explain why rhetorical questions do not violate felicity
conditions even though they are not interpreted as ordinary
information-seeking questions: They are used in indirect speech acts, which
explains why they do not seek information, and in such speech acts, questions
are evaluated against the common ground. Rhetorical questions thus emerge a
means of presenting a statement not as the speaker's personal opinion, but as
a consequence of the common ground, which explains their persuasive effect.
It is a central observation of discourse theory that readers or
hearers interpret discourse as a coherent whole, even though aspects
of the interpretation are left implicit in the text. Specifically,
they fill in content relations and rhetorical links between discourse
units when no explicit discourse particles or connectives are present.
For instance, in
(1) Max switched off the light.
The room became dark.
readers normally assume that there is a causal relation
linking the two propositions: the room became dark because Max
switched off the light.
Theories of discourse interpretation such as RST (MannThompson 1988),
LTAG for discourse (Webber 2004), and LDM (Polanyi 1988), and SDRT
(Asher and Lascarides 2003) account for this on the assumption that
the reader's linguistic knowledge contains a set of discourse
(coherence, rhetorical) relations. Reasoning on the basis of his world knowledge and an
linguistic analysis of what is explicitly stated in the discourse, the
reader will select, or infer, an appropriate discourse relation in
cases linking the propositions expressed in (1). Discourse relations
are a necessary ingredient in accounting for coherence. Not only do
they provide missing or underspecified information, they also
contribute to the hierarchical structure of discourse. While each of
the above mentioned approaches presupposes a different type of
structure for discourse, all of them claim that the hierarchical
structure of the discourse context constrains the interpretation of
incoming sentences or utterances.
In this talk I want to review some of the motivation for the latter
assumption, and zoom in on the function and formal treatment of
discourse relations within what might be called 'grammatical'
approaches to discourse analysis such as the above ones. I will make
use of 'LDG for Discourse', a framework for the formal specification
of discourse theory developed in Van Leusen (2007). Due to its
neutral, constraint-based set-up, it is specifically tuned the
comparison and synthesis of insights from different approaches.
Moreover, it supports a rich treatment of discourse analysis by
integrating syntactic, semantic, and pragmatic constraints and world
knowledge in a single reasoning system. Among the issues to be
addressed are the default inferential character of discourse
relations, comparability with the semantics and pragmatics of overt
discourse operators, and the not so clear link with speech act theory.
If time allows, I will ponder on the possibility of eliminating
discourse relations as such from the theory.
The DRT semantics of aspectual adverbs is applied to their effects in dialogue, conditionals and interrogative contexts. Their presuppositions may serve to constrain quantifiers in subject position and interesting forms of linguistic variety in gradual quantification is discussed. The temporal meaning of aspectual adverbs blends into meta-linguistic uses where generic background constraints may or may not support the current context. The DRT tools are claimed to provide sufficient possibilities to account for the observed inferences, obviating the need for special notions of default inference or normal worlds which are hard to give precise content.
H. Smessaert & A. ter Meulen (2004), Temporal reasoning with aspectual adverbs,
Linguistics and Philosophy 27.2, 209-262.
Recently, the notion of scalar discourse relations has come up. E.g., Asher and Lascarides (2003) argue that certain continuations to discourses are better than others, concluding that, say, the Narration relation is scalar and a stronger Narration assigned to the better
continuation.
The examples discussed in the literature all concern coordinating relations. My focus will be on subordinating relations, specifically Elaboration. I will argue that there are two senses in which a
discourse unit can be more or less elaborating a superordinate discourse unit: a qualitative sense (the one employed by Asher and Lascarides (2003)) and a structural sense that can be derived from structural distance.
I will discuss some of the interactions between these two senses, suggest a measure for the degree of informativeness of an elaboration and discuss the usefulness of this measure for deciding whether an elaboration is relevant. I will argue that the notion of relevance as
introduced by Sperber and Wilson (1986) is not adequate for the problem at hand and suggest a different line to pursue.
Currently, we face a wide range of theories of NPI licensing (logic, syntactic, implicature-based, scale based). While some are explicitely tailored for specific types of NPIs, others leave the range of intended applications implicit and restrict discussion to the prototypical cases (any, ever). Krifka, Heim and others suggest that NPIs which are licensed by hidden "even" are strong NPIs. On basis of sentences with overt "even", I will argue that hidden "even" is not what causes an NPI to be strong. I will then propose that Antiveridicality plus scalar licensing can account for a wide variety of NPI data, including strong and weak cases.I will present an account of Condition B effects, Strong Crossover effects, and certain puzzling restrictions on the interpretation of pronouns in VP-ellipsis.