ZASPiL 60 = Proceedings of Sinn und Bedeutung 22, Vol. 1
Refine
Document Type
- Part of a Book (29)
Language
- English (29)
Has Fulltext
- yes (29)
Is part of the Bibliography
- no (29)
Keywords
- focus (3)
- presupposition projection (2)
- presuppositions (2)
- pronoun (2)
- quantification (2)
- quantifiers (2)
- scalar implicature (2)
- subjectivity (2)
- tense (2)
- uniqueness (2)
- Epistemic Containment Principle (ECP) (1)
- Korean (1)
- MaxElide (1)
- Q-adverbs (1)
- Question Under Discussion (QUD) (1)
- Thai (1)
- Tiwa (1)
- accessibility (1)
- acquisition (1)
- adjectival antonyms (1)
- adjectives (1)
- agreement mismatch (1)
- alternative questions (1)
- alternative semantics (1)
- appositives (1)
- aspect (1)
- assertion (1)
- assertions (1)
- at-issue content (1)
- binding (1)
- choice functions (1)
- classifiers (1)
- coercions (1)
- coherence relations (1)
- common ground (1)
- comparative constructions (1)
- conditionals (1)
- conventional implicatures (1)
- conversational implicatures (1)
- cornering (1)
- counterfactuals (1)
- de dicto (1)
- decomposition, (1)
- definiteness (1)
- deixis (1)
- deontic modals (1)
- depiction verbs (1)
- determiners (1)
- discourse (1)
- discourse coherence (1)
- discourse particles (1)
- distributional semantics (1)
- domain restriction (1)
- donkey sentences (1)
- double access (1)
- ellipsis (1)
- embedded implicature (1)
- entailment (1)
- epistemic 'modals' (1)
- epistemic indefinites (1)
- epistemic modals (1)
- event semantics (1)
- events (1)
- evidentiality (1)
- exhaustivity (1)
- experimental pragmatics (1)
- experimental semantics (1)
- experiments (1)
- familiarity (1)
- firsthand experience (1)
- frame semantics (1)
- free choice (1)
- function words (1)
- future (1)
- game–theoretic pragmatics (1)
- generic quantifier (1)
- gestures (1)
- habitual (1)
- habituals (1)
- hierarchy (1)
- higher-order quantification (1)
- imperfective (1)
- implicated presupposition (1)
- implicature (1)
- imposters (1)
- individual variation (1)
- inferencing task (1)
- information structure (1)
- intensional quantifiers (1)
- intensional transitives (1)
- knowledge (1)
- language acquisition (1)
- lexical semantics (1)
- linear order (1)
- local context (1)
- logical form (1)
- manner implicature (1)
- maximality (1)
- mention-some (1)
- miners puzzle (1)
- modal flavor (1)
- modification (1)
- monotonicity (1)
- natural language metaphysics (1)
- negation (1)
- negative strengthening (1)
- negative-islands (1)
- non-intersective adjectives (1)
- not-at-issue content (1)
- number construction (1)
- parsing (1)
- partitives (1)
- person agreement (1)
- personal reference (1)
- perspective (1)
- pragmatic inference (1)
- predicates of personal taste (1)
- probability (1)
- progressive (1)
- prominence (1)
- pronouns (1)
- psycholinguistics (1)
- quantifier processing (1)
- reaction time (1)
- reasoning errors (1)
- reconstruction (1)
- relational adjectives (1)
- relative clauses (1)
- resumptive pronouns (1)
- salience (1)
- scalar diversity (1)
- scalar implicatures (1)
- scalar inferences (1)
- scope (1)
- secondary focus (1)
- similarity (1)
- simplification (1)
- speech reports (1)
- speeded verification (1)
- syllogisms (1)
- type composition logic (1)
- type-shifting (1)
- unalternative semantics (1)
- visusal representations (1)
- wide scope indefinites (1)
This paper is about what Ninan (2014) (following Wollheim 1980) calls the Acquaintance Inference (AI): a firsthand experience requirement imposed by several subjective expressions such as Predicates of Personal Taste (PPTs) (delicious). In general, one is entitled to calling something delicious only upon having tried it. This requirement can be lifted, disappearing in scope of elements that we will call obviators. The paper investigates the patterns of AI obviation for PPTs and similar constructions (e.g., psych predicates and subjective attitudes). We show that the cross-constructional variation in when acquaintance requirements can be obviated presents challenges for previous accounts of the AI (Pearson 2013, Ninan 2014). In place of these, we argue for the existence of two kinds of acquaintance content: (i) that of bare PPTs; and (ii) that of psych predicates, subjective attitudes and overt experiencer PPTs.
For (i), we propose that the AI arises from an evidential restriction that is dependent on a parameter of interpretation which obviators update. For (ii), we argue that the AI is a classic presupposition. We model both (i) and (ii) using von Fintel and Gillies’s (2010) framework for directness and thus connect two strands of research: that on PPTs and that on epistemic modals. Both phenomena are sensitive to a broad direct-indirect distinction, and analyzing them along similar lines can help shed light on how natural language conceptualizes evidence in general.
Imposters are grammatically third-person expressions used to refer to the firstperson speaker or second-person addressee (e.g. ‘the present authors’ when used to refer to the first-person writer, ‘Mommy’ or ‘Daddy’ when used by parents for self-reference in child-directed speech). Current analyses of imposters differ in whether they derive the unusual referential properties of imposters using syntactic means or attribute them to semantic and pragmatics. We aim to shed light on these competing approaches by means of a psycholinguistic experiment focusing on first-person imposters that investigates the kinds of pronouns (first-person vs. third-person) used to refer to imposter antecedents. Our results show that manipulating the prominence of the first-person speaker does not significantly boost the acceptability of first-person pronouns in imposter-referring contexts. However, our results suggest that a purely syntactic approach may not be sufficient either, as psycholinguistic processing factors also appear to be relevant.
The Epistemic Containment Principle (ECP) requires that epistemic modals take wider scope than strong quantifiers such as every or most (von Fintel and Iatridou, 2003). Although fairly robust in its realization, a few systemic classes of counterexamples to the ECP have been noted. Based on these, previous work has argued for two claims: subjective modals obey the ECP, whereas objective ones don’t (Tancredi, 2007; Anand and Hacquard, 2008); and every respects the ECP, whereas each violates it (Tancredi, 2007). This paper argues that explicit Questions Under Discussion (QUDs; Roberts, 1996; Ginzburg, 1996) also systematically influence the ECP: scopal orderings that provide relevant answers to the given QUDs are preferred, and this tendency can override the ECP. To support this claim, the paper presents an experimental study. The results corroborate the existence of systematic QUD effects on the ECP, and support the view that the ECP is derived from a confluence of various pragmatic and lexical biases.
This paper compares the modal particle fei (Schlieben-Lange, 1979; Thoma, 2009) with the modal particle/sentence adverb aber (not to be confused with the conjunction aber, ‘but’). Intuitively, both items express some form of contrast and correction. We will show that both are special among discourse particles in the following sense: They make a contribution that is interpreted at a level distinct from the level where at-issue content (Potts, 2005) is interpreted, as is standard for modal particles (see Gutzmann, 2015 and the references therein). But more interestingly, they exclusively relate to propositions that have not entered the Common Ground via being the at-issue content of an assertion made by the addressee.
Experimental studies investigating logical reasoning performance show very high error rates of up to 80% and more. Previous research identified scalar inferences of the sentences of logical arguments as a major error source. We present new analytical tools to quantify the impact of scalar inferences on syllogistic reasoning. Our proposal builds on a new classification of Aristotelian syllogisms and a closely linked classification of reasoning behaviors/strategies.
We argue that the variation in error rates across syllogistic reasoning tasks is in part due to individual variation: reasoners follow different reasoning strategies and these strategies play out differently for syllogisms of different classes.
We bring experimental considerations to bear on the structure of comparatives and on our understanding of how quantifiers are processed. At issue are mismatches between the standard view of quantifier processing cost and results from speeded verification experiments with comparative quantifiers. We build our case in several steps:
1. We show that the standard view, which attributes processing cost to the verification process, accounts for some aspects of the data, but fails to cover the main effect of monotonicity on measured behavior. We derive a prediction of this view for comparatives, and show that it is not borne out.
2. We consider potential reasons - experimental and theoretical - for this theory-data mismatch.
3. We describe a new processing experiment with comparative quantifiers, designed to address the experimental concerns. Its results still point to the inadequacy of the standard view.
4. We review the semantics of comparative constructions and their potential processing implications. 5. We revise the definition of quantifier processing cost and tie it to the number of Downward Entailing (DE) operators at Logical Form (LF). We show how this definition successfully reconciles the theory-data mismatch. 6. The emerging picture calls for a distinction between the complexity of verified representations and the complexity of the verification process itself.
In this paper I argue for a new constraint on questions, namely that a question denotation (a set of propositions) must map to a partition of a Stalnakerian Context-Set by point-wise exhaustification (point-wise application of the function Exh). The presupposition that Dayal attributes to an Answer operator follows from this constraint, if we assume a fairly standard definition of Exh (Krifka, 1995). But the constraint is more restrictive thereby deriving the sensitivity of higher order quantification to negative islands (Spector, 2008).
Moreover, when combined with recent proposals about the nature of Exh - designed primarily to account for the conjunctive interpretation of disjunction (e.g. Bar-Lev and Fox, 2017) - Dayal’s presupposition follows only in certain environments. This observation allows for an account of the "mention-some" interpretation of questions that makes specific distributional predictions.
This paper is an attempt to systematically investigate how contrastive focus interacts with various types of not-at-issue content (co-speech and post-speech gestures, lexical presuppositions, and appositives). I look, in particular, at when focus forces at-issue interpretations of typically not-at-issue content, when it does not, and when such at-issue interpretations are impossible even to satisfy focus-related requirements.
I conclude that the main factors affecting how a given type of content aligns along these dimensions are its prosodic (in)dependence and level of attachment in the syntax. The two factors also interact in a non-trivial way, in particular for gestures, which I use as a basis for an analysis of gestures that does not assume that their temporal alignment directly determines their semantics (contra Ebert and Ebert, 2014; Ebert, 2017; Schlenker, 2018), but instead relies on syntax/semantics and syntax/prosody interaction.
I present data that suggest the universal entailments of counterfactual donkey sentences aren’t as universal as some have claimed. I argue that this favors the strategy of attributing these entailments to a special property of the similarity ordering on worlds provided by some contexts, rather than to a semantically encoded sensitivity to assignment.
Tiwa (Tibeto-Burman; India) has two series of epistemic indefinites: one whose epistemic effects arise via an anti-singleton constraint similar to Spanish alg´un (Alonso-Ovalle and Men´endez-Benito, 2010), and another, wide-scope indefinite whose epistemic effects must be derived differently. I propose that for these latter indefinites, ignorance arises not through domain constraints, but as a result of their choice functional nature through competition with other indefinites. Tiwa’s wide scope indefinites then constitute a new kind of epistemic indefinite, showing that ignorance implicatures for indefinites can arise through different sorts of competition.
This paper experimentally investigates presupposition projection from the scope of the quantifiers every and at least one, as triggered by the factive verb be aware and the iterative adverb again.
The first issue we are concerned with is whether presuppositions project universally or existentially from quantified sentences. Different theoretical accounts endorse opposing views here (e.g., Heim, 1983; Geurts, 1999; Beaver, 2001; Schlenker, 2008, 2009; Fox, 2012), while recent experimental work (Chemla, 2009; Tiemann, 2014) suggests that the force of the projected presupposition varies by quantifier.
The second issue we look at is how the descriptively observed readings arise—in particular, as a direct result output from the projection mechanism, or via additional, independent mechanisms such as domain restriction (e.g., Geurts and van Tiel, 2016): if the domain of the quantifier is restricted, this can yield what looks like non-universal inferences in light of the overall, unrestricted domain, even if the projection mechanism itself yields a universal presupposition. Finally, we test whether the presupposed content also forms part of the entailed content, at least for certain triggers (Sudo, 2012; Klinedinst, 2016; Zehr and Schwarz, 2016).
Our results yield clearly different patterns for every and at least one, with every giving rise to universal presuppositions, which, to a very limited extent, can be weakened by domain restriction, and at least one overwhelmingly giving rise to non-universal presuppositions. Our results also indicate the availability of presupposition-less readings for both triggers in the task at hand, apparently more prevalent than domain restriction. Thereby, we present novel evidence that helps to pinpoint which of the theoretical options can be substantiated experimentally.
This paper argues that extant approaches to presupposition projection that either rely on strict linear order (Schlenker, 2009) or hierarchy (Romoli and Mandelkern, 2017) cannot provide a uniform account of data drawn from head-final languages. While building on Schlenker’s theory, this paper resolves the issues by restricting the calculation of local contexts to specific points in the parsing process. The consequence is that the theory makes a prediction robust to the head directionality parameter
Reference of pronouns may be constrained via lexical presuppositions, including marked F-features, implicated presuppositions, and deictic center shifting in certain languages.
This paper explores the acquisition of personal reference terms in Thai, a language that has a highly complex personal reference system. The participants of the study were 67 typicallydeveloping children (TD) and 29 children with autism spectrum disorders (ASD), a population which has long been observed to have difficulties with pronouns. The children were asked to complete simple production and comprehension tasks on personal reference terms. Overall, ASD children performed on par in production but significantly poorer in comprehension than TD children. Given the freedom of choice in the production task, ASD children preferred using fixed referential terms for self-reference, whereas TD children opted for personal pronouns. In terms of comprehension, ASD children were shown to generally be able to detect the person features but they seemed to struggle the most with the pragmatic aspects of personal reference terms. Our results support previous literature that lexical presuppositions are acquired earlier than implicated presuppositions. We add to the literature that the types or the amount of implicated presuppositions matter in acquisition
Depiction verbs such as paint license i(mage)- and p(ortrait)-readings; for instance, Ben painted a cow can convey that Ben produced an image of an unspecific cow or a portrait of a specific cow. This paper takes issue with a property-based intensional analysis of depiction verbs (Zimmermann, 2006b, 2016) and instead argues for an extensional account. Accordingly, the i-reading is rooted in the introduction of worldly representations by the explicit noun cow as such, whereas the p-reading is rooted in the interpolation of an implicit representation via coercion. This take on the ambiguity captures the following key traits. On i-readings, only representations are accessible to quantifiers and anaphors; moreover, intensional effects such as substitution failure disappear once ordinary objects and representations are adequately distinguished. P-readings, by contrast, involve representations that depend on the portrayed ordinary objects as particulars; correspondingly, only ordinary objects are accessible to quantifiers and anaphors. The proposal is spelled out in Asher’s (2011) Type Composition Logic.
We present an experiment which tests children’s comprehension of the requirements of use of pronouns and definites. An adult-like use of definites and pronouns imposes different but related requirements. In the case of definites, a unique referent is required in the context, whereas in the case of a pronoun, the referent in the context has to be salient. In this experiment, we use a novel word task to test three-year-olds’ sensitivity to these requirements. Our results show that children are adult-like in their sensitivity to salience in their comprehension of pronouns, compared to definites. However, they failed to show sensitivity to the uniqueness requirement on the use of definites.
Miners and modals
(2018)
I generalise Kolodny and MacFarlane’s miners puzzle by showing epistemic analogues of their case exist. After motivating various conservative approaches to the original problem, I show how they fail to solve the problem in its epistemic guise. I argue that a probabilistic approach to information-sensitivity gives a general solution to the problem.
The Wolof imperfective auxiliary di is compatible with event-in-progress, habitual and future readings. Furthermore, while varieties of all these readings are available for di when it sits in a syntactically low position, only future readings are available when it sits in a syntactically high position. We aim to account for this puzzle by combining several ingredients independently motivated in the literature: (i) event-relative circumstantial modality for event-in-progress, habitual, and a subset of future readings; (ii) metaphyisical modality for generalized future readings; (iii) the idea that syntactic height determines the type of modal anchor that projects a modal base. This study contributes to our understanding of the relation between syntactic height and modal flavor, as well as the nature of modal-aspectual interactions cross-linguistically.
Previous research on scalar implicature has primarily relied on metalinguistic judgment tasks and found varying rates of such inferences depending on the nature of the task and contextual manipulations. This paper introduces a novel interactive paradigm involving both a production and a comprehension component, thereby fixing a precise conversational context.
The main research question is what is reliably communicated by some in this communicative setting, when the quantifier occurs in unembedded positions as well as embedded positions. Our new paradigm involves an action-based task from which participants’ interpretation of utterances can be inferred. It incorporates a game–theoretic design, including a precise model to predict participants’ behaviour in the experimental context.
Our study shows that embedded and unembedded implicatures are reliably communicated by some. We propose two cognitive principles which describe what can be left unsaid. In our experimental context, a production strategy based on these principles is more efficient (with equal communicative success and shorter utterances) than a strategy based on literal descriptions.
In recent years, experimental research has demontrated great variability in the rates of scalar inferences across different triggering expressions (Doran et al. 2009, 2012, van Tiel et al. 2016). These studies have been taken as evidence against the so-called uniformity assumption, which posits that scalar implicature is triggered by a single mechanism and that the behaviour of one scale should generalize to the whole family of scales. In the following, we present an experimental study that tests negative strengthening for a variety of strong scalar terms, following up on van Tiel et al. (2016). For example, we tested whether the statement John is not brilliant is strengthened to mean that John is not intelligent (see especially Horn 1989). We show that endorsement rates of the scalar implicature (e.g., John is intelligent but not brilliant) are anti-correlated with endorsements of negative strengthening. Further, we demonstrate that a modified version of the uniformity hypothesis taking into account negative strengthening is consistent with van Tiel et al.’s data. Therefore, variation across scales may be more systematic than suggested by the van Tiel et al. study.
Alternative Questions with "or not" (NAQ) convey a cornering effect, which is not found with they polar counterparts (PQ). This effect has been claimed to consist of two parts (Biezma 2009): NAQs (i) cannot be used discourse-initially and (ii) they do not license followup questions/subquestions.
In this paper, we ask the following: Are both parts of cornering linked to the same property of NAQs? Or do they reflect distinct linguistic phenomena? We explore the issue by comparing the behavior of NAQs to Complement Alternative Questions (CAQ), a type of question that, like NAQs, presents logically opposite alternatives but, unlike NAQs, fully spells out the second one. Results from two experiments suggest that both parts of cornering can instead be explained in terms of independent semantic and pragmatic principles, which operate beyond the domain of alternative questions.
Sentences containing subjective predicates - e.g., "The movie was awesome"” - are intuitively anchored to a particular perspective; this makes them different from sentences describing objective facts - e.g., "The movie was set in 1995".
While authors have long debated on whether this intuition tracks a lexical distinction between subjective and factual predicates, much remains to be explored on whether, and how, the difference between these two assertions is reflected at the illocutionary level. Relying on evidence from two experiments, we show that assertions containing subjective predicates display different discourse behavior from objective assertions. We take these findings to support the idea that SAs should be assigned a special illocutionary profile, unveiling a genuine empirical difference between subjective and factual speech.
Analyses of scope reconstruction typically fall into two competing approaches: 'semantic reconstruction', which derives non-surface scope using semantic mechanisms, and 'syntactic reconstruction', which derives it by positing additional syntactic representations at the level of Logical Form. Grosu and Krifka (2007) proposed a semantic-reconstruction analysis for relative clauses like the gifted mathematician that Dan claims he is, in which the relative head NP can be interpreted in the scope of a lower intensional quantifier. Their analysis relies on type-shifting the relative head into a predicate of functions. We develop an alternative analysis for such relative clauses that replaces type-shifting with syntactic reconstruction. The competing analyses diverge in their predictions regarding scope possibilities in head-external relative clauses. We use Hebrew resumptive pronouns, which disambiguate a relative clause in favor of the head-external structure, to show that the prediction of syntactic reconstruction is correct. This result suggests that certain type-shifting operations are not made available by Universal Grammar.
According to Ogihara (1995), the usage of the embedded present in a speech report such as John said that Mary is in the room is restricted by the cause of John’s belief (the state that made John think that Mary is in the room): the present tense can be used only if this cause still holds at the time that John said that Mary is in the room is uttered.
This paper presents experimental evidence demonstrating that this is only one of the factors that licenses a felicitous usage of the embedded present tense. In particular, we show that the cause of belief still holding is not a necessary condition, and identify two additional, sufficient (but not necessary) factors: in cases of false belief, who is aware of the falsity of the belief and duration of the reported state. While these factors are independent, they collectively support the idea that the present tense encodes ‘current relevance’, even in embedded contexts (e.g. Costa 1972; McGilvray 1974). This gives rise to the question of how we can derive ‘current relevance’ and, in particular, whether previous analyses of the embedded present tense are adequately equipped to do so.
This paper presents a new account of the generalization that focused elements cannot be elided, framed within Unalternative Semantics, a framework that does away with syntactic F-marking. We propose the mirror image of the generalization: what is elided cannot introduce alternatives. We implement this as a focus restriction in UAS and then go on to show how to account for MAXELIDE effects using the same technique, without making reference to any transderivational constraints.
The semantics of adjectives related to nominals denoting societal roles, such as presidential (from president), have remained understudied. We examine the semantics of what we call role-denoting relational adjectives, providing a formal analysis using the notion of a frame, a unified representation for lexical knowledge, world knowledge, and context. The frames we propose are based on a constructivist philosophical understanding of social roles, leading us to posit a multi-tiered ontology of events and individuals. Using frames and our ontology, we provide a general semantics for role-denoting relational adjectives and roles
Discourses in the historical (or narrative) use of the simple present in English prohibit backshifting, though they allow forward sequencing. Unlike both reference time theories and discourse coherence theories of these temporal inferences, we propose that backshifting has a different source from narrative progression. In particular, we argue that backshifting arises through anaphora to a salient event in the preceding discourse.
The goal of this paper is to evaluate two approaches to quantification in event semantics, namely the analysis of quantificational DPs in terms of generalized quantifiers and the analysis proposed in Schein (1993) according to which quantifiers over individuals contain an existential quantifier over sub-events in their scope. Both analyses capture the fact that the event quantifier always takes scope under quantifiers over individuals (the Event Type Principle in Landman (2000)), but the sub-events analysis has also been argued to be able to account for some further data, namely for adverbs qualifying ‘ensemble’ events and for mixed cumulative/ distributive readings. This paper shows that the sub-events analysis also provides a better account of the Event Type Principle if a broader range of data is considered, including cases with non-existential quantifiers over events: unlike the generalized quantifiers analysis, it can successfully account for the interpretation of indefinites in bare habituals and sentences that contain overt adverbs of quantification.
Korean is a generalized classifier language where classifiers are required for numerals to combine with nominals. This paper presents a number construction where the classifier is absent and the numeral appears prenominally. This construction, which I call the classifier-less number construction (Cl-less NC), results in a definite or a partitive reading where the referent must be familiar: ‘the two women’ or ‘two of the women’. In order to account for this, I argue that Korean postnominal number constructions are ambiguous between a plain number construction and a partitive construction. After motivating and proposing an analysis for the partitive structure, I argue that Cl-less NC is derived from the partitive construction, explaining its distributional restriction and the interpretation.
Counter to the often assumed division of labour between content and function words, we argue that both types of words have lexical content in addition to their logical content. We propose that the difference between the two types of words is a difference in degree. We conducted a preliminary study of quantificational determiners with methods from Distributional Semantics, a computational approach to natural language semantics. Our findings have implications both for distributional and formal semantics. For distributional semantics, they indicate a possible avenue that can be used to tap into the meaning of function words. For formal semantics, they bring into light the context-sensitive, lexical aspects of function words that can be recovered from the data even when these aspects are not overtly marked. Such pervasive context-sensitivity has profound implications for how we think about meaning in natural language.