Proceedings of Sinn und Bedeutung
Not a member yet
1139 research outputs found
Sort by
Long-distance cumulativity asymmetry: experimental evidence from Czech
Get PDFThis paper explores conjunctions and the phenomenon of cumulative asymmetry with respect to the subject and object position. The tested conjunction is the Czech \u27i\u27, which is postulated to be a D-conjunction. The aim of this paper and the experiment carried out is to test whether Czech speakers observe cumulativity asymmetry at long-distance in ECM constructions and whether monoclausal and long-distance configurations differ in interpretations
Conversational dynamics of Russian questions with \u27razve\u27
No full textRussian questions with the particle \u27razve\u27 appear to convey negative bias in conflicting evidence scenarios, thus bearing superficial resemblance to English questions with \u27really\u27 (cf. Repp and Geist forth.). I argue that \u27razve\u27-questions convey a novel type of bias and signal that the speaker is in a situation with belief revision potential, facing a conflict between a prior belief and a current abductive inference. Depending on context, such questions receive (i) an information-seeking interpretation or (ii) what I will call a ‘point-making’ interpretation. I propose a unified semantics for \u27razve\u27, while also showcasing the limitations of current approaches to question bias and making a case for sensitivity to abduction in a novel empirical domain
An Algebra of Thought that predicts key aspects of language structure
Get PDFThe Meaning First Approach hypothesizes that humans can form complex non-linguistic representations in an ‘Algebra of Thought’ independent of any language used in communication. Since the Algebra of Thought and language nevertheless must be related, one research program is to reverse engineer the Algebra of Thought from what is known about language. In this paper, we focus on universal structural properties of human languages. We investigate an Algebra of Thought fragment containing logical conjunction, a part-whole relationship and two cognitive efficiency requirements that exclude redundancies. We show that at least three universal structural properties of languages follow from these assumptions: cartographic hierarchies, the obligatory decomposition of non symmetric binary predicates, and the obligatory lexical content of dependent elements in binding dependencies
Cumulative readings of distributive conjunctions: Evidence from Czech and German
Get PDFWe present new data showing that cross-linguistically a class of conjunction strategies usually considered purely distributive exhibit cumulative readings in object position. This is similar to the pattern found for every-DPs (Kratzer 2003; Champollion 2010 a.o.) and German jed-DPs (Haslinger and Schmitt, 2018), suggesting all these elements form a natural class. Our evidence stems from experiments on Czech A i B and German sowohl A als auch B (‘A as well as B’) using a Semantic Choice Task (Lohiniva and Panizza 2016). In the crucial items, participants saw pictures of a cumulative scenario and a scenario making both the cumulative and the distributive reading false, and had the option of selecting one scenario or rejecting both. In both languages, cumulative scenarios were accepted more often with the conjunction in object position than with the conjunction in subject position. Further, as surface subjects of passive sentences patterned with objects of active sentences and topicalized objects with non- topicalized objects, passivization and topicalization do not seem to affect cumulative readings
Putting plural definites into context
Get PDFTheories of plural definites differ with respect to ‘non-maximal readings’ of plural definites in positive and negative sentences. The implicature approach predicts an inherent asymmetry where plural definites allow for non-maximal readings in positive sentences but not in negative ones. The non-implicature approach makes symmetric predictions that non-maximal readings should be available to the same degree in positive and negative sentences. Previous experimental work found evidence for an asymmetry between positive and negative cases, but since they did not control for potential contextual effects, it remains to be a possibility that positive and negative sentences were judged against different implicit contexts that had different effects on the availability of non-maximal readings. In this paper, we report on two experiments using a picture-sentence verification task, testing the effect of context on the non-maximal readings of plural definites in positive and negative sentences. We tested sentences containing the plural definite ‘his/her presents’ under a positive quantifier ‘every boy/girl’ and two negative quantifiers ‘no boy/girl’ and ‘not every boy/girl,’ manipulating what was relevant in the context. Our results indicate that while non-maximal readings under all three quantifiers are modulated by context, the effect size is smaller for no than for the other two quantifiers. We argue that these findings pose challenges for both types of theories, and discuss possible amendments of each approach in order to account for our findings
About ‘Us’: Clusivity ⊔ exh
Get PDFThis paper argues that two generalizations about person pronouns and agreement point towards a semantic account where both exhaustification and cumulative coordination can apply word-internally. Both generalizations involve reference to pluralities that include both the author and the addressee for which many languages use an inclusive first person morph distinct from exclusive first person and second person. The first generalization (originally due to Zwicky 1977, CLS Proceedings) holds that in terms of person categories, in languages that lack an inclusive/exclusive contrast the inclusive neutralizes with the first person and never the second person, despite overlapping meaning with both. The second generalization, which we argue is exemplified by Mandarin and Daur, is that in languages which mark clusivity only in some subsystems, but not others, then not only will the inclusive meaning be expressed as a first (not second) person, but where there are common forms between the (sub)systems, then it will be the exclusive, not the inclusive that expresses the general first person in environments that lack a clusivity contrast. Our account assumes only the two person features AUTHOR and PARTICIPANT, but uses exhaustivization instead of feature negation or Maximize Presuppostion and assumes that feature conjunction can be cumulative coordination. By analyzing the first person inclusive as AUTHOR cumulatively conjoined with exhaustivized PARTICIPANT, we derive the two generalizations noted
\u27Czech\u27 the alternatives: A probe recognition study of focus and word order
No full textComprehenders have been found to activate, select, and represent plausible alternatives to focused elements when processing incoming sentences (see Gotzner & Spalek, 2019 for an overview). This is consistent with Rooth’s (1992) theory of focus interpretation, which claims that the function of focus is to create an additional level of meaning consisting of a set of propositions derived by replacing the focused element with its contextually appropriate alternatives of the same semantic type. However, the psycholinguistic research on the processing of focus has mostly been done on a small sample of Germanic languages which mostly use prosody to mark focus. We tested whether the current results generalise to Czech, which can use word order to mark narrow focus. We report on a probe recognition study aiming to test whether Czech comprehenders represent alternatives to focused subjects. The results provide preliminary evidence in favour of this claim
Alternatives and jurisdiction in predication
Get PDFWhile many predicates can compose consistently (e.g. \u27This dog is happy\u27), some can only compose via conjunctive material like additive particles (e.g. \u27This comedy is #(also) a tragedy\u27). This paper asks what relation must exist between predicates for them to fall in the latter category. In previous work, I suggested that predicates require \u27also\u27 if they come from the same conceptual taxonomy. In the present paper, I show that another factor is at play, namely whether two predicates contribute the same kind of information (‘have the same jurisdiction’) in a given sentence. In particular, same-taxonomy predicates stop requiring \u27also\u27 when they are interpreted with a different jurisdiction. From the observation that jurisdiction is pertinent to whether \u27also\u27 is required, I suggest that jurisdictional identity is in fact the only factor in whether two expressions require an additive; bringing in the notion of taxonomic co-membership is superfluous
On the context dependence of artifact noun interpretation
Get PDFThe context dependence of artifact noun category boundaries is underexplored relative to interpretational context dependence in other linguistic domains (e.g., gradable adjectives). Taking inspiration from a normative debate over the role of context in legal interpretation (in particular, the role of legislators’ policy goals), we show experimentally that contextual information as to a rule’s purpose systematically modulates interpreter beliefs about the category boundaries of artifact nouns contained within the rule. We propose a Bayesian pragmatic model of the context-dependent resolution of artifact noun extensions, which we compare against context-independent baselines. Our experimental and modeling results suggest the need for an explicitly context-sensitive, multi-dimensional degree semantics for artifact nouns
Reasoning about stored representations in semantics using the typology of lexicalized quantifiers
Get PDFThe typology of lexicalizations in natural languages is highly skewed: some meanings repeatedly receive their own expression as individual morphemes or words in language after language, while many other meanings rarely or never do. For example, while many languages have monomorphemic counterparts of English some and all, no known language has a monomorphemic quantifier that means ‘all or none’ or a quantifier that asserts that its two arguments are of the same cardinality. It seems tempting to reason from this typological skew to properties of stored representations. However, it is not generally safe to assume that if something is typologically unattested then it simply cannot be represented or learned. The representational system for stored denotations is just one of several interacting factors that affect the typology, and other factors such as communicative pressure and learnability are likely to shape patterns of lexicalization. In this paper we propose to reason from the typology to stored representations by modeling the representational framework, communicative pressure, and learnability directly within an evolutionary model, building on work by Brochhagen et al. (2018). Our empirical focus is a lexicalization asymmetry noted by Horn (1972) in the domain of logical operators and framed within the Aristotelian Square of Opposition. We show that, on certain assumptions, Horn’s lexicalization pattern depends on very particular representational costs in the lexicon: it arises if the storage costs for ‘every’ and ‘some’ are lower than those for ‘not every’ and ‘not some’ but not otherwise