Univ.-Prof. Dr. Hana Filip, PhD
Univ.-Prof. Dr. Jacopo Romoli, PhD
Lehrstuhl Semantik
Institute for Language and Information (Institut für Sprache und Information)
Heinrich Heine University
Wednesday 10:30 a.m. – 12:00 p.m.
Building 23.21 / Room 04.23
The meetings are also streamed via Webex. Please use the following Webex link to participate remotely: https://hhu.webex.com/meet/tsnider
November 30, 2022
Wednesday, 10:30-12:00
23.21.04.23 in person
Online: https://hhu.webex.com/meet/tsnider
Jacopo Romoli
Universität Düsseldorf
“What makes an inference robust?”
Sentences like (1) embedding disjunction in the scope of an existential or universal modal strongly suggest that first year students can take introduction to logic and that they can take semantics.
(1) It is permitted/required that first year students take introduction to logic or semantics.
This inference is generally referred to as ‘free choice’ in the case of existential modals, and ‘distributive’ or ‘diversity’ in the case of universal ones. A prominent approach treats both of these inferences as scalar implicatures. This proposal, however, faces well-known challenges coming from experimental results showing that the two inferences are generally more robust, faster to process, and easier to acquire than regular scalar implicatures. A common response to this challenge is to hypothesize that such discrepancies stem from the type of alternatives used to derive them. This talk reports on a series of experiments that tested this hypothesis by comparing positive, disjunctive sentences like (1) to variants of these sentences involving negation and conjunction, for which the implicature approach predicts similar inferences on the basis of the same type of alternatives. Our results reveal that, while the two inferences are indeed quite robust in the disjunctive cases, the inferences that their negative, conjunctive variants give rise to are not. These findings are challenging for the hypothesis that the type of alternatives involved in implicature computation is responsible for differences in robustness. And more generally are challenging for any unified account of these inferences, whether based on implicature or not. We end by outlining a hybrid account treating the less robust inferences as implicatures, and the more robust ones as entailments. All in all, comparing inferences in terms of robustness is a powerful perspective to learn more about the nature of those inferences and the relation between them.
——————————————————————————————————————————-
November 16, 2022
Wednesday, 10:30-12:00
23.21.04.23 in person
Online: https://hhu.webex.com/meet/tsnider
Yichi (Raven) Zhang
Universität Düsseldorf
“Epistemic Pseudo-Agreement and the Hyper-Domain Semantics for Epistemic Modals”
The literature on epistemic modals often makes use of group knowledge to specify a relevant body of information as the quantificational base for modals. Standardly, distributed knowledge is taken as the obvious candidate for this purpose. In this project, we argue that using distributed knowledge as the quantificational base over-generates in cases of epistemic pseudo-agreement: it incorrectly predicts disagreement about “might p” while intuitively no such disagreement would arise. To tackle this issue, instead of seeking a more sophisticated way to pool information from discourse participants, we propose a hyper-domain semantics where we use a hyper-domain, a set of information states each representing a discourse participant’s private information, to serve as the relevant body of information for modals.
Ascending one level on the set-theoretic hierarchy also allows us to address certain inadequacy of the traditional domain semantics (Yalcin 2007) regarding the embedding of modals under the knowledge operator. To further investigate the corresponding logic of our hyper-domain semantics, we take inspiration from Schulz’s (2010) translation of the informational consequence in domain semantics into the modal logic S5 and devise a translation of the consequence relation in our hyper-domain semantics into a new logic called “Logic of Epistemic Disagreement” interpreted over possibility models (cf. Holliday 2021).
——————————————————————————————————————————-
November 9, 2022
Wednesday, 10:30-12:00
23.21.04.23 in person
Online: https://hhu.webex.com/meet/tsnider
Patrick D. Elliott
Joint work with Yasutada Sudo (University College London)
Universität Düsseldorf
“Anaphora and Simplification”
In the literature on exhaustification, inferences associated with complex sentences are
derived by strengthening relative to simpler alternatives. Consider a disjunctive sentence of the form [p or q] — much recent work makes crucial use of so-called domain alternatives p and q. A prominent example is the case of free choice: in order to explain the fact that [may p or q] implies may p and may q, exhaustification accounts strengthen the meaning of the modalized disjunction relative to alternatives [may p] and [may q] (see, e.g., Bar-Lev and Fox 2020). In this talk, we identify a fundamental problem for accounts of such simplification-based inferences: a discourse-anaphoric dependency is possible between the simpler alternatives, as in the example “You may include no appendix, or typeset it in Palatino”. The challenge is how to derive the inference ‘you may include an appendix typeset in Palatino’, given that the second disjunct is an open sentence. As a proof of concept, we account for this inference by extending Goldstein’s semantic account of free choice within the framework of Elliott’s (2022) Bilateral Update Semantics. Ultimately, we’ll show that this problem is much more general, and arises for many such simplification-based inferences, such as distributive inferences (Crnič et al 2015), and simplification of disjunctive antecedents in conditionals (Nute 1975).
——————————————————————————————————————————-
November 2, 2022
Wednesday 10:30 a.m. – 12:00 p.m.
Lisa Hofmann
University of Stuttgart
Anaphora and Negation
The talk presents a commitment-based dynamic analysis of anaphoric accessibility, and experimental data that supports a commitment-based approach to anaphoric polarity sensitivity. I take this to be an argument about the kind of information that a language user needs to store in memory when interpreting a discourse. Specifically, I argue that the constraints placed on anaphoric relations by negation and other non-veridical operators, can be explained in a meaningful way by keeping track of speaker commitments about discourse referents (drefs). This provides a unified account of various phenomena that have previously received disparate analyses in the literature.
I present formal and experimental work from my dissertation (Hofmann, 2022) ‘Anaphora and Negation’, which suggests that anaphoric accessibility can be understood more broadly as sensitivity to a combination of three factors: (i) the veridicality of the local intensional context of the antecedent, (ii) the veridicality of the local intensional context of the anaphor, and (iii) the relationship between these two contexts.
As a result, constraints on anaphoric accessibility are captured as constraints on relations between intensional drefs. More specifically, I suggest that assuming intensional representations of drefs, which can be related to an intensional representation of speaker commitments, allows for an account that is both parsimonious and explanatory. This is so because it derives the constraints on anaphoric accessibility based on two basic principles of discourse interpretation: (i) the referent of an individual denoting phrase exists in its local intensional context, and (ii) a discourse should contain consistent (non-contradictory) information.
——————————————————————————————————————————-
October 19, 2022
Wednesday 10:30 a.m. – 12:00 p.m.
Kurt Erbach
Heinrich Heine University & University of Würzburg
The Development of Object Mass Nouns
Hybrid format
In person: 23.21.04.23
Virtually: https://hhu.webex.com/meet/tsnider
While research has shown that users of typologically distinct languages distinguish substances and objects regardless of their countability system (cf. Japanese and English in Imai & Gentner 1997), and while this distinction can in some way be seen in grammatical patterns that one say reveals mass nouns and count nouns (e.g. Chierchia 2021), it seems that English stands apart in having a much larger set of nouns that bisect these two categories, such as object mass nouns (Erbach 2021). What remains an open question from the point of view of evolutionary linguistics is what conditions allowed for English to have so many object mass nouns while languages like Hungarian, Greek, and Japanese have so few, if any at all in the case of the latter two (see Erbach 2021). To shed light on this question, I present diachronic data from English that suggests Old English might not have had object mass nouns, rather that the category developed and expanded greatly in the Middle English period due to a large number of borrowings and reanalysis of morphology, giving rise to the particularly productive —ware and -wear morphemes. This is taken to suggest that the productivity of derivational morphology is a key factor in the amount of object nouns in a given language. To probe whether there are grammatical constraints that prevent languages like Greek and Japanese from having object mass nouns as assumed by Chierchia (2019), I will also propose several experiments that could be used to test whether novel object mass nouns could be learned in these languages.
——————————————————————————————————————————-
July 13, 2022
10:30 a.m. – 12.00 p.m.
Building 24.21 / Room 05.61 (Hybrid Meeting)
This meeting will also be streamed via Webex. Please use the following Webex link to participate remotely: https://hhu.webex.com/meet/tsnider
Sonia Ramotowska
Heinrich Heine University
Quantifying quantifier representations with the Diffusion Decision
Model: The case of most and more than half
The Generalized Quantifier Theory (GQT; Mostowski 1957; Barwise & Cooper, 1981) provides a formal framework to express the meaning of quantifiers in the form of truth conditions. In this talk, I will present a case study of two broadly discussed natural language quantifiers – most and more than half (e.g., Hackl, 2009; Solt, 2016). The GQT predicts that most and more than half are truth conditionally equivalent. However, the growing body of evidence shows substantial differences in processing (e.g., Hackl, 2009) and pragmatics (Solt, 2016; Carcassi & Szymanik, 2021) of these quantifiers. A satisfactory explanation of these differences is still missing. I will present data from two online behavioral experiments in which I directly tested the prediction of the GQT. In these experiments, participants verified a sentence with a quantifier (e.g., most and more than half) against a proportion given as a percentage (e.g., 60%). I will argue that computational modeling is essential to investigating the truth-conditional representations of quantifiers. I will show how different aspects of meaning can be modeled by using the state-of-the-art mathematical model of decision-making – the Diffusion Decision Model (Ratcliff, 1978). The results of the experiments challenge the GQT assumption. Moreover, they call for a more psychologically realistic treatment of semantic representations that would account for vagueness and individual differences in meanings.
——————————————————————————————————————————-
July 6, 2022
10:30 a.m. – 12.00 p.m.
Building 24.21 / Room 05.61 (Hybrid Meeting)
This meeting will also be streamed via Webex. Please use the following Webex link to participate remotely: https://hhu.webex.com/meet/tsnider
Paolo Santorio
University of Maryland, College Park
The Modal Perfective and Actuality Entailments
In several languages, a number of modal expressions give rise to actuality entailments when combined with perfective aspect: “pfv-modal [p]” entails p. Several accounts of this phenomenon have been proposed, but they all suffer from serious empirical shortcomings. I argue that actuality entailments motivate adopting a modal theory of the perfective. In particular, I suggest that the perfective works as a kind of “settledness” moda: “pfv [P(e)]” states that the P-event e holds at exactly the same time intervals at all worlds in a historical modal base. This new meaning entails the standard meaning of the perfective (with some empirical advantages). In addition, in combination with plausible assumptions, it predicts actuality entailments for circumstantial and ability modals.
——————————————————————————————————————————-
June 29, 2022
10:30 a.m. – 12:00 p.m.
This meeting will be streamed via Webex. Please use the following Webex link to participate remotely: https://hhu.webex.com/meet/tsnider
Paul Marty
L-Università ta’ Malta/UCL
Disjunction introduction: logical failure or pragmatic success?
Research on human reasoning has identified various cases where our capacity to reason seems prone to systematic ‘failures’. These failures can in principle stem from two distinct sources. Human reasoning might simply not follow the normative standards of classical-logical validity (a.o., Tversky & Kahneman 1974, Braine & O’Brien 1998, Walsh & Johnson-Laird, 2004). Alternatively, what looks like reasoning failures might instead be the result of perfectly sound reasoning acting on implicit, yet predictable enriched interpretations of the premises and/or conclusion (a.o., Mascarenhas & Koralus 2017, Mascarenhas & Picat 2019). In this talk, I report on a series of experiments investigating the interplay between reasoning proper and interpretive processes using inferences to disjunctive-like elements (e.g., concluding John visited Paris or London from John visited London). In their basic form, these deductive problems involve disjunctive conclusions which are robustly rejected by naïve subjects despite being classically valid. Results evidence that the driving force behind people’s rejection of such conclusions is in fact a multiplicity of conflicting pragmatic inferences. On the theoretical side, I will argue that these findings align with the predictions from state-of-the-art theories of scalar implicatures and further demonstrate the role of pragmatic considerations in standard reasoning tasks. On the methodological side, I will discuss novel paradigms permitting to attenuate the impact of pragmatics in such tasks, making people in effect `more logical’.
——————————————————————————————————————————-
June 22, 2022
10:30 a.m. – 12:00 p.m.
Location: Building 24.21 Room 05.61
This meeting will also be streamed via Webex. Please use the following Webex link to participate remotely: https://hhu.webex.com/meet/tsnider
Wataru Uegaki
University of Edinburgh
(presenting joint work with Mora Maldonado & Jennifer Culbertson)
Learnability and constraints on the semantics of clause-embedding predicates
Responsive predicates are clause-embedding predicates like English “know” and “guess” that can take both declarative and interrogative clausal complements. The meanings of responsive predicates when they take a declarative complement and when they take an interrogative complement are hypothesized to be constrained in systematic ways across languages, suggesting that these constraints represent semantic universals (Spector & Egre 2015; Theiler et al. 2019; Uegaki 2019; Roelofsen & Uegaki 2020). Following previous studies that have investigated links between learning and semantic universals (Chemla et al. 2019; Steinert-Threlkeld & Szymanik, 2020; Saratsli, et al. 2020; Maldonado & Culbertson, 2021), we tested whether semantic constraints on responsive predicates are reflected in learning. Our results suggest that adults learning a novel clause-embedding predicate in the lab infer a constraint (clausal distributivity, in particular) without explicit evidence.
——————————————————————————————————————————-
June 15, 2022
10:30 a.m. – 12:00 p.m.
Location: Building 24.21 Room 05.61
This meeting will also be streamed via Webex. Please use the following Webex link to participate remotely: https://hhu.webex.com/meet/tsnider
Jakub Szymanik
University of Amsterdam
Universal preference for convex concepts
Natural languages prefer to lexicalize convex meanings both in content and function domains. For example, color terms across languages denote convex regions of color space, and languages tend to lexicalize only convex quantifiers. In the talk, I’ll argue that such bias can be explained cognitively by simplicity. In particular, I will show that convex quantifiers have shorter minimal-description lengths and lower Kolmogorov complexity. I will also present computational and behavioral experiments showing that convex concepts, quantifiers and colors, are easier to learn. Finally, I will give a computational simulation of iterated learning, suggesting that individual learning biases may add up in the process of evolution, making convex convexity an emergent property.
——————————————————————————————————————————-
June 8, 2022
no talk, SALT 32 conference
——————————————————————————————————————————-
June 1, 2022
online: https://hhu.webex.com/meet/tsnider
Jon Ander Mendia
Universitat Autònoma de Barcelona
Nominal reference beyond individuals
In some contexts DPs have the ability to refer outside the nominal domain, denoting instead amounts or degrees rather than individuals: while (1a) tells us something about four individual pizzas, (1b) tells us about ‘an amount of four pizzas’.
1a) Four pizzas are vegetarian.
1b) Four pizzas is enough.
Examples like these have been used to argue in favor of a so-called degree/individual polysemy, where the locus of this variation lies in the dual senses of the nominals themselves (Brasoveanu 2009, Rett 2014). This talk explores the syntactic and semantic properties of these DPs and concludes that, although they may well be individual/degree polysemy in the grammar, this may not be the best explanation for the behavior of (1b).
——————————————————————————————————————————-
May 25, 2022
on campus: Building 24.21 / 2421.05.61
online: https://hhu.webex.com/meet/tsnider
10:30 – 12:00 UTC+2 (Germany)
Agata Renans
Ruhr-Universität Bochum
(joint work with Anne Mucha)
Actuality entailments and aspect
There is a general agreement in the literature that verbs have an argument slot for an
event argument in their structure (Davidson 1967, Carlson 1984, Krifka 1992,
Maienborn 2008, 2019, Filip 2000, 2005, among many others). There is also an
agreement that the event argument introduced by the verb must be bound at some
point in the structure. However, there is an ongoing discussion on where and how this
happens. In this talk, I will present experimental work on ‘Actuality Entailments’ in
German which potentially sheds light on that issue.
——————————————————————————————————————————-
May 18, 2022
on campus: Building 24.21 / 2421.05.61
online: https://hhu.webex.com/meet/tsnider
10:30 – 12:00 UTC+2 (Germany)
Clemens Mayr
University of Göttingen
(joint work with Ekaterina Vostrikova)
A unified semantics for exceptive-additive ‘besides’
This talk is about the semantics of exceptive-additive ‘besides’ in English. Similarly to exceptive constructions (like ‘except’ and ‘but’), `besides’ can occur with universal quantifiers and it contributes the negative inference in such contexts (as shown in (1)).
(1) Every girl except/ but/besides Ann came. (Inference: Ann did not come)
Unlike exceptives, ‘besides’ can occur with existentials and in such contexts it contributes a positive inference (as the contrast between (2) and (3) illustrates).
(2) *At least one/ exactly one/ some girl(s) girl except/ but Ann came.
(3) At least one/ exactly one/ some girl(s) besides Ann came. (Inference: Ann came)
In this talk we propose a unified semantic treatment of exceptive-additive ‘besides’ that accounts for its interaction with various quantifiers and explains its difference from exceptives like ‘but’. The account we suggest is based on independently motivated mechanisms. Specifically, we propose to extend the existing analyses of exceptives in terms of Exh (Hirsch 2016; Črnic 2021) to exceptive-additives. The crucial difference between the two types of constructions lies in the way the alternatives are constructed.
——————————————————————————————————————————-
May 11, 2022
online: https://hhu.webex.com/meet/tsnider
Natasha Korotkova
University of Konstanz
Hearsay and (non-)commitment
Evidentiality has figured prominently in theorizing about speech acts and evidential expressions have been argued to update discourse commitments of the interlocutors, signalling that the speaker—in root declaratives—commits to having a particular type of evidence for φ and, oftentimes, to φ itself. Such utterances have been treated as special assertions (e.g., weak, modalized, or hedged) except for utterances with hearsay evidentials. Across languages (AnderBois 2014), hearsay evidentials often do not require the speaker to commit to φ, which is diagnosed by disavowals ‘ φ is not the case’ or ‘I don’t believe φ’. Because assertion requires belief/commitment, utterances with hearsay evidentials have been argued to be speech acts of presentation whose goal is to raise a particular issue. In this talk, I provide a novel account of non-commitment with hearsay wherein such utterances are assertions whose at-issue contribution is the speaker’s having hearsay evidence. This proposal is rooted in the idea that the evidential contribution has a variable at-issue status. It is not-at-issue by default but can also become at-issue, which is what happens with disavowals. Utterances with hearsay evidentials can therefore be treated as ordinary assertions, just like corresponding speech reports of the form ‘I hear that φ’. This account makes the prediction that only expressions with a variable at-issue status would allow non-commitment. The prediction is borne out for slifting parentheticals, which, unlike evidentials, have a fixed not-at-issue status.
——————————————————————————————————————————-
May 4, 2022
online: https://hhu.webex.com/meet/tsnider
Peter Sutton
Universitat Pompeu Fabra, Barcelona
Polysemy and Copredication: A situation-theoretic approach
This talk presents a new situation-theoretic approach to polysemous expressions such as lunch, book, and statement. The noun book, for example, is polysemous between a physical entity reading (1a) and an informational entity reading (1b), and it also licenses copredication across these readings (1c).
(1) a. the thick book b. the interesting book c. the thick and interesting book
In order to account for polysemy and copredication, Asher and Pustejovsky (2013, among others) assume a bespoke dot-type constructor such that (1c) can refer to one object that has different ‘aspects’ such that the dot type reflects the different aspects of this object. Gotham (2017) assumes that (1c) can refer to one object that is a mereological sum of a physical object and an informational object which requires, contrary to standard mereological assumptions, that sums can be formed across denotational domains.
In contrast, I argue that we do not need to assume that sentences such as (1c) can refer to one ‘object’. Adopting an assumption common from situation theoretic semantics, and specifically from Type Theory with Records (TTR) (e.g., Cooper 2022), I claim that polysemous nouns denote situations that contain entities. Importantly, it is unsurprising situations can be complex and so witness multiple entities of different types, and so no polysemy specific types or mereological sums need to be assumed. In short, I claim that polysemous nouns denote situations that contain multiple entities and the lexical semantics of the noun specifies how these different entities are related. For example, (1c) denotes a situation that contains one physical book and one informational entity (its contents) such that the meaning of book specifies that the latter is the contents of the former. Moreover, I propose that it is lexically specified relations such as Contents, Theme etc. that license copredication. To take another example, lunch is eventuality-physical entity polysemous and the copredication in (2) is licensed because the physical food stuff is the Patient of the lunch-eating eventuality.
(2) Lunch lasted two hours and was delicious (adapted from Asher and Pustejovsky 2013)
An advantage of this approach is that it can explain restrictions on copredication for more complex cases of polysemy than book. For example, statement is three-ways polysemous between a physical entity reading (Phys), an informational entity reading (Inf) and an eventuality reading (Ev). However, not all combinations of these readings are available for copredication as we see in (3b) and (5a).
(3) a. The statement in the envelope is inaccurate. (Phys, Inf)
b. ?The statement in the envelope lasted half an hour. (Phys, Ev)
(4) a. The inaccurate statement lasted half an hour. (Inf, Ev)
b. The inaccurate statement was sealed in an envelope. (Inf, Phys)
(5) a. ?The half-hour statement was sealed in an envelope. (Ev, Phys)
b. The half-hour statement was inaccurate. (Ev, Inf)
The account provided in this talk straightforwardly accounts for these patterns. The markedness of (3b) and (5a) is predicted because there is a contents relation between physical statements and their informational contents and between stating eventualities and their informational contents, but no relation that is specified by the meaning of statement holds between the eventuality and the physical object.
——————————————————————————————————————————-
April 20, 2022
online: https://hhu.webex.com/meet/tsnider
Jacopo Romoli
Heinrich Heine Universität Düsseldorf
Presenting joint work with Paul Marty, Yasu Sudo, and Richard Breheny (UCL)
Implicating in semi-cooperative contexts
In ordinary conversations, disjunctive sentences like June visited Frankfurt or Düsseldorf are commonly understood as conveying that she didn’t visit both cities (exclusivity), and that the speaker doesn’t know which of the two cities she visited (ignorance) (Grice 1975, Gazdar 1979, Horn 1972 a.o.). There is general consensus that these inferences are not conveyed as part of the literal meaning, but rather they arise as implicatures. On the standard pragmatic approach, implicatures are the output of implicit reasoning on the part of the hearer over why the speaker said what she said instead of something else (Grice 1975, Horn 1972, Gazdar 1979, Sauerland 2004, Geurts 2010, Chemla 2010, van Rooij & Schulz 2004 among others). This approach derives exclusivity and ignorance on the basis of the same assumptions and reasoning. This predicts that these two inferences should either be both associated with the sentence or neither of them should. This prediction has recently been argued to be challenged by so-called ‘semi-cooperative’ contexts; contexts where it is presupposed that the speaker is not going to provide all relevant information that they have (e.g. a game show, a treasure hunt etc) (Fox 2014). The claim has recently received some experimental support (Agyemang 2020). Taking Agyemang’s experimental findings as our starting point, the study I report in this talk has three aims. Firstly, we aimed to test an alternative explanation for these findings, which derives the exclusivity implication not via any mechanism for scalar implicature but via background knowledge. We ran an experiment that control for such background knowledge and found that exclusivity emerges to the same extent, thus confirming and strengthening the challenge. Secondly, we extended the investigation to presuppositional sentences like Bill is not aware that June visited Frankfurt or Düsseldorf, which have been argued to give rise to the same exclusivity and ignorance inferences (Spector & Sudo 2017, Marty & Romoli 2021b a.o.). In this case as well we find that presupposed exclusivity emerges robustly even in the absence of presupposed ignorance, thus extending the challenge for the pragmatic approach at the presuppositional level. Finally, we outline a more nuanced pragmatic derivation of inferences related to disjunction which can explain how ignorance and exclusivity arises in normal cooperative contexts, while only exclusivity arises in these semi-cooperative ones.
——————————————————————————————————————————-
February 9, 2022
Yael Greenberg
(Joint work with Micky Daniels)
Ben-Gurion University of the Negev
What does even even do with extreme-adjectives comparatives?
Extreme adjectives (e.g. gigantic, terrible, gorgeous) were observed in the literature to be degraded in comparative constructions relative to their non-extreme counterparts (e.g. big, bad, pretty). However, the presence of the focus sensitive particle even in them was shown to improve this degraded status. Examples of this pattern are (1) and (2), from Morzycki (2012):
1. ? Godzila is more gigantic than Mothra. (cf. Godzila is bigger than Mothra)
2. Godzila is even more gigantic than Mothra.
In this talk we report work in progress where we ask what it is precisely about even which leads to this improving effect, and where we suggest that the answer to this question can contribute to a debate which exists in the literature concerning the nature of the scalar presupposition of this particle. We first examine an analysis of the improving effect based on the widely-used unlikelihood-based scalar presupposition of even. We point out, however, a number of issues for this analysis. Following these issues we turn to develop an explanation based on an alternative, gradablity-based, characterization of the scalar presupposition of even and argue that it fares better than the unlikelihood-based one. In particular, we rely on the ‘mapped-scale’ component in the gradability-based scalar presupposition, and suggest that it leads to an indirect domain-widening effect of even, which in turn overcomes the difficulty of comparing degrees on the non-salient portion of the scale, argued to be associated with extreme adjectives. (This analysis differs from an early attempt of ours – based on Maximize-Presupposition! – to explain the pattern in (2)). We support our present analysis by showing that other constructions which were independently argued to involve mapped scales improve the status of EA-comparatives as well.
November 24, 2021
11:30-13:00 UTC+1
David Erschler
Ben-Gurion University of the Negev
Colloquial emphatic negation in Russian and morphology of negative concord
The now widely adopted approach to the negative concord (NC) of Zeijlstra (2004) leaves much freedom in how to analyze the morphology of negative indefinites. In particular, the analysis of the Slavic ni-negative indefinites (e.g. the Russian ni-kto NI-who ‘no one’; ni-čego NI-what ‘nothing’ etc.) remains a matter of debate. In this talk, I explore the properties of an emphatic negation construction from a highly colloquial register of Russian. Alongside the standard negation, colloquial Russian has grammaticalized an alternative negation marker, xuj ‘dick’ and its euphemisms, that does not license ni-words. Using the properties of this construction, I provide novel evidence that morphological feature copying is implicated in forming ni-negative indefinites in Slavic.
November 17, 2021
Seth Cable
University of Massachusetts Amherst
Two Paths to Habituality: The Semantics of ‘Habitual Mode’ in Tlingit
Handout: http://hanafilip.com/wp-content/uploads/2021/11/TlingitHab-Duesseldorf.pdf
In this talk, I will present a detailed description and formal semantic analysis of habitual sentences in Tlingit (Na-Dene; Alaska, British Columbia, Yukon). As in many other languages (Carlson 2005, 2012), there are two means in Tlingit for expressing a habitual statement, such as ‘my father eats salmon’. The first employs a relatively unmarked verb, realizing imperfective aspect. In the second type of habitual sentence, however, the verb bears special ‘habitual’ morphology. Although there is a significant overlap in the use of these constructions, certain semantic contrasts do exist. Most notably, the special habitual marking cannot be used to express pure, unrealized ‘dispositions’/‘functions’/‘duties’ (e.g., ‘Mary handles any mail from Antarctica’). In other words, Tlingit habitual morphology – unlike imperfective aspect – requires the ‘habituality’ in question to have actually occurred, an effect that has also observed for habitual morphology in a variety of other, unrelated languages (Green 2000, Bittner 2008, Boneh & Doron 2008, Filip 2018). I develop and defend a formal semantic analysis that captures these (and other) contrasts between imperfective and habitual verbs. In brief, imperfective aspect is argued to possess a modal semantics, quantifying over alternative worlds/situations (Arregui et al. 2014, Ferreira 2016). Habitual morphology, however, is argued to be associated with a (potentially covert) quantificational adverb, one that quantifies strictly over times in the actual world. The consequences of this account for the analysis of habitual sentences in other languages are explored. Most notably, we find that (i) ‘habituality’ so-called is potentially a heterogeneous phenomenon, and resists unified definition or semantic analysis, and (ii) therefore is a ‘sui generis’ category of phenomena, which cannot be reduced as an instance of aspect or modality (Filip & Carlson 1997, Filip 2018).
November 10, 2021
Luuk Suurmeijer
Heinrich-Heine Universität Düsseldorf
Compositionality in Distributional Formal Semantics
Compositionality—the construction of complex meaning from simpler constituents—is a fundamental notion in the study of linguistic meaning. Two prominent approaches to modelling meaning are Formal Semantics and Distributional Semantics. While in Formal Semantics, compositionality takes center stage, there is no straightorward means to represent graded probabilistic phenomena. Distributional Semantics, by contrast, is inherently probabilistic, but does not adequately capture truth-conditional constructions such as quantification.
A novel framework for meaning representation, Distributional Formal Semantics (DFS), combines the strengths of both approaches, by offering graded probabilistic meaning representations in the form of vectors that are truth-conditionally grounded. This allows DFS to model logical and probabilistic phenomena, and to remaincompositional on the propositional level. While it has been shown that meaning representations in DFS can be incrementally constructed using an RNN, I here aim to provide an explicit, set-theoretic mechanism of compositional meaning construction that derives propositional meaning representations from their subpropositional constituents. By arriving at such a mechanism we scale up the coverage of DFS from the confined, small-scale training data of an RNN to the full compositional complexity of natural language.
| November 3, 2021 Jesús Calvillo Tinoco Heinrich-Heine Universität Düsseldorf Connectionist Language Production: Distributed Representations and the Uniform Information Density Hypothesis We approach the modeling of human sentence production using artificial recurrent neural networks and distributed semantic representations. The main questions we focus on are: (i) whether the distributed semantic representations defined by Frank et al. (2009) are suitable to model sentence production generalizing systematically, (ii) the behavior and internal mechanism of a model that uses this representations and (iii) a mechanistic account of the Uniform Information Density Hypothesis (UID; Jaeger, 2006; Levy and Jaeger, 2007). Regarding the first point, the semantic representations of Frank et al. (2009), called situation vectors are points in a vector space where each vector contains information about the observations in which an event and a corresponding sentence are true. These representations have been successfully used to model language comprehension (e.g., Frank et al., 2009; Venhuizen et al., 2018). These representations were used as input for a sentence production model that implements an extension of a Simple Recurrent Neural network (Elman, 1990), testing the model under different conditions corresponding to different levels of systematicity, also known as compositional generalization. Systematicity is an essential aƩribute that a model of sentence processing has to possess, considering that the number of sentences that can be generated for a given language is infinite, and therefore it is not feasible to memorize all possible message-sentence pairs. The results showed that the model was able to generalize with a very high performance in all test conditions, demonstrating a systematic behavior. Regarding the second point, the sentence production model was analyzed in two different ways. First, by looking at the sentences it produces, including the errors elicited, revealing that the model learns the syntactic patterns of the language, reflecting its statistical nature, and that its main difficulty is related to very similar semantic representations, sometimes producing unintended sentences that are however very semantically related to the intended ones. Second, the connection weights and activation paƩerns of the model were also analyzed, reaching an algorithmic account of the internal processing of the model. Regarding the third point, an extension of the model is proposed with the goal of modeling UID. According to UID, language production is an efficient process affected by a tendency to produce linguistic units distributing the information as uniformly as possible and close to the capacity of the communication channel, given the encoding possibilities of the language, thus optimizing the amount of information that is transmitted per time unit. This extension of the model approaches UID by balancing two different production strategies: one where the model produces the word with highest probability given the semantics and the previously produced words, and another one where the model produces the word that would minimize the sentence length given the semantic representation and the previously produced words. |
| July 7, 2021 12:30 – 2 pm, online via Webex https://hhu.webex.com/meet/tsnider Julie Hunter Linagora Relations and Structure in Discourse Interpretation In this talk we consider a variety of ways that the semantic relations that we infer between propositional level contents—and the often complex structures to which they contribute—can influence semantic interpretation. In particular, we show how recent work on discourse relations and structure has been used to model phenomena along the semantics-pragmatics interface, to analyze multimodal discourse, and to provide an account of discourse goals and the interaction of bias and discourse interpretation. |
| June 16, 2021 17:00-18:30, online via Webex https://hhu.webex.com/meet/tsnider Andrea Beltrama University of Pennsylvania Imprecision and speaker identity: How social meaning affects pragmatic reasoning Recent work at the interface of semantics and sociolinguistics proposed that the linguistic signaling and uptake of social identity traits can be captured via models similar to those used to formalize pragmatic inferences (Acton 2019; Burnett 2019; Henderson and McCready 2020) and non-at-issue content (Smith et al. 2010); and provided empirical evidence that listeners reason about the semantic/pragmatic properties of linguistic utterances to draw social inferences about the speaker (Beltrama 2018; Jeong 2018). These findings raise the question of whether reverse effects exist as well, i.e., whether (and how) social meanings can also impact the interpretation of semantic/ pragmatic meanings. We provide experimental evidence that (i) numerals receive stricter interpretations when uttered by Nerdy (vs. Chill) speakers; and that (ii) this effect is stronger for comprehenders who donմ (strongly) identify with the speaker, suggesting that pragmatic reasoning is crucially shaped by social information about both the speaker and the comprehender. Taken together, these findings suggest that social meanings are more intricately intertwined with pragmatic reasoning than previously thought: not only is the signaling and uptake of social meanings amenable to being formalized through the same reasoning patterns that can be applied to pragmatic inferences; but social meanings are also recruited by comprehenders to resolve the indeterminacy surrounding the interpretation of semantic meanings. The emerging picture is one in which different layers of meanings inform one other in a bi-directional fashion, highlighting the importance of integrating the study of social content into theoretical and experimental semantics and pragmatics. |
| June 9, 2021 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Moshe E. Bar-Lev Language, Logic & Cognition Center (LLCC) The Hebrew University of Jerusalem Obligatory implicatures and the relevance of contradictions Magri (2009, 2011) proposed a theory of some cases of oddness based on an obligatory calculation of implicatures. While conceptually motivated, his theory has been shown to have limited empirical coverage. Alternative proposals improved on the empirical side, but lacked conceptual motivation. The goal of this talk is to explore a modification of Magri’s view which preserves the conceptual motivation while achieving wider empirical coverage. On this view oddness comes about due to a conspiracy of two assumptions: (i) in the computation of implicatures, alternatives cannot be ignored if their presence leads to a relevant meaning (BarLev 2018); (ii) contradictions are relevant (following Lewis 1988). |
| May 12, 2021 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Ashwini Deo The Ohio State University A strong question: Unifying exclusivity, mirativity, precisification, and intensification in Marathi =c Several Indo-Aryan languages, including Bengali, Gujarati, Hindi, and Marathi contain a discourse clitic whose uses overlap with those of English discourse adverbials like the exclusives only/just, intensifiers really/absolutely/totally, precisifiers right/exactly, expectation confirmers indeed/that very/exactly, and scalar additive even, without corresponding perfectly to any of them. I offer a unified analysis of the varied and seemingly disparate uses of this expression type, focusing on the Marathi variant =c. Under the analysis I will discuss, =c signals that an underspecified, vague question is being interpreted so as to maximize the informativity of the answer provided by the prejacent. The analysis treats =c as a generalized ҩnformativity booster” that comments on the strength of the question that the prejacent resolves. This discourse function proposed for =c is substantively different from the discourse function associated with exclusives (like only), which are said to comment on the strength of the prejacent among alternative answers to a contextually fixed question. The analysis proposed here thus identifies a new possible function for discourse management expressions, enriching our understanding of the typology of such expressions. Within Marathi, this new perspective allows us to treat =c as the pragmatic antonym of another enclitic particle in Marathi =tar which is analyzed as signaling that the question resolved by its prejacent is weak. At the crosslinguistic level, this new perspective opens the door to a better understanding of discourse particles by offering an explanation for why exclusivity, mirativity, precisification, and intensification often cluster together in many languages, another example I will consider being English just. |
| May 5, 2021 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Lelia Glass Georgia Institute of Technology Quantifying relational nouns in corpora While relational nouns (cousin) are traditionally delineated in a binary and theory-dependent manner, this paper approximates relationality as a continuous, objective corpus metric (Percent Possessive) – allowing for lexicon-wide exploration of which nouns are more or less relational and why. Comparing across nouns and accounting for the ontological class of the nounճ referent, I find that Percent Possessive is positively correlated with a nounճ per-million-word frequency. Comparing across different web communities, I find that a noun is more frequent, and shows a greater ratio of definite to indefinite tokens, in the community where its Percent Possessive is significantly higher. I take these findings to be consistent with the claim that a noun is more relational (as measured by Percent Possessive) when human interaction with its referent is more conventional (as measured by its frequency and definite-to-indefinite ratio). Inspired by the many authors who have suggested a socio-cultural component to relationality and possession, this paper explores at scale in English how nouns reflect the conventions of the people who use them. Talk slides |
| April 28, 2021 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Noortje Venhuizen Universitt des Saarlandes Distributional Formal Semantics Formal Semantics and Distributional Semantics offer complementary strengths in capturing the meaning of natural language. As such, a considerable amount of research has sought to unify them, either by augmenting formal semantic systems with a distributional component, or by defining a formal system on top of distributed representations. Arriving at such a unified formalism has, however, proven extremely challenging. One reason for this is that formal and distributional semantics operate on a fundamentally different ‘representational currency’: formal semantics defines meaning in terms of models of the world, whereas distributional semantics defines meaning in terms of linguistic context. An alternative approach from cognitive science, however, proposes a vector space model that defines meaning in a distributed manner relative to the state of the world. In this talk, we will present a re-conceptualization of this approach by deriving it from well-known principles of formal semantics, thereby demonstrating its full logical capacity. The resulting Distributional Formal Semantics is shown to offer the best of both worlds: contextualized distributed representations that are also inherently compositional and probabilistic. We will then show how these representations can be used for ‘deep language modeling’, by introducing a neural network model that captures various semantic phenomena, including probabilistic inference and entailment, negation, quantification, reference resolution and presupposition. Talk slides |
| January 20, 2021 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Kurt Erbach und Leda Berio Universitt Bonn and HHU Dsseldorf Counting and categorizing: The relationship between the mass/count distinction and thought In this talk, we propose that the count/mass distinction has meddler and spotlight effects on thought (see Wolff and Holmes 2010) which is supported by the interaction between syntax and categorization in noun acquisition of speakers of languages with distinct counting morphosyntax (e.g. Gordon 1985; Imai & Gentner 1997; Lucy & Gaskins 2003). It has been shown that speakers of languages with different counting systems both make use of the same cognitive tools, albeit differently (Inagaki & Barner 2009; Barner, Inagaki & Li 2009). For example, in English counting constructions, nouns are distinguished as either count or mass while in Japanese counting constructions, all nouns are counted with classifiers. Despite these differences, speakers of both languages behave one way with respect to nouns that refer to complex objects and another way with nouns that refer to non-solid substances (Inagaki & Barner 2009, Imai & Gentner 1997) though there are some differences across languages in behavior with nouns that refer to simple objects and non solid substances (Imai & Gentner 1997; cf. Lucy & Gaskins 2003). While it has been argued that this state of affairs rules out stronger Whorfian accounts of language (Bale & Barner 2019), we provide a more detailed account of the relation between the count/mass distinction and thought. Talk slides |
| January 13, 2021 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Nina Haslinger Georg-August-Universitt Gttingen Plurals as higher-order scalar predicates: Unifying cumulativity and non-maximality Sentences like (i) are traditionally taken have distinct cumulative and distributive readings. While the distributive reading requires Ann and Sarah to each like all of the four neighbors, the cumulative reading merely requires that Ann and Sarah each like at least one of the neighbors and each neighbor is liked by at least one of them. This is usually viewed as an ambiguity that reflects the presence or absence of distributivity or cumulation operators in the LF of such sentences (for an overview, see Champollion to appear). (i) Ann and Sarah like their four neighbors. In this talk, I present English and German data in support of two empirical claims that complicate this standard picture. First, the availability of cumulative readings appears to be a context-dependent matter: A sentence containing two or more plurals is acceptable in a cumulative scenario only if that scenario answers the QUD or ԩssueՠraised in the discourse context in the same way as a distributive scenario. I argue that this contextdependency does not just reflect pragmatic pressure to address the QUD. Instead, it seems to be of a piece with a semantic constraint defended by Kriz (2015) for non-maximal readings of definites (see also Malamud 2012): Roughly, a sentence like (ii) may be true in a non-maximal scenario only if that scenario provides the same answer to the QUD/թssueՠas a maximal scenario. (ii) The neighbors are asleep. Second, the acceptability of sentences like (i) in cumulative scenarios is influenced by the extensions of alternative predicates: (i) can clearly be judged true if Ann likes neighbors 1 and 2 and Sarah likes neighbors 3 and 4, but seems less acceptable if the scenario further specifies that Ann absolutely hates 3 and 4 and Sarah absolutely hates 1 and 2. Again, this behavior parallels previous observations about standard cases of nonmaximality like (ii), where it seems to make a difference whether the `exceptional’ neighbors are reading quietly or having a noisy street party (cf. Kriz 2015). In sum, there are far-reaching parallels between cumulative readings of sentences with multiple plurals and nonmaximal readings of plural definites. Yet, since these parallels can be replicated with cumulative readings of predicate and sentential conjunctions ((iii), cf. Schmitt 2019) and cumulative readings of quantifiers ((iv), cf. Schein 1993 a.o.), cumulativity cannot be reduced to a non-maximal interpretation of the individual plurals involved. (iii) The ten experts said that Macron will resign and that Merkel won’t resign. (iv) Ann and Sarah read every book on the list. I show how this data pattern can be captured within a nonstandard plural semantics that relies both on the analogy between plural quantification and vague scalar predication (Burnett 2017) and on the idea that complex constituents containing a plural expression themselves receive a plural denotation via special composition rules (Schmitt 2019). Plural expressions of any category denote partially ordered scales with a ositive setՠand a Ԯegative setՊ that parallels the positive and negative extension of vague scalar predicates. In particular, like the scales for vague predicates, these plural scales may have a gap between the positive and the negative set which is resolved in a QUD-dependent manner, accounting for the context-dependency of cumulative and non-maximal interpretations. Further, the analogy between plural predication and scalar predication gives rise to a unified account of the role of alternative predicates in the semantics of cumulative and non-maximal sentences. References Burnett, Heather (2017): Gradability in natural language. Logical and grammatical foundations (Oxford Studies in Semantics and Pragmatics 7). Oxford University Press. Champollion, Lucas (to appear): Distributivity, collectivity, and cumulativity. In Daniel Gutzmann, Lisa Matthewson, Ccile Meier, Hotze Rullmann & Thomas Ede Zimmermann (eds.), Blackwell Companion to Semantics. Kriz, Manuel (2015): Aspects of homogeneity in the semantics of natural language. University of Vienna dissertation. Malamud, Sophia A. (2012): The meaning of plural definites: A decision-theoretic approach. Semantics and Pragmatics 5(3). 1–58. Schein, Barry (1993): Plurals and Events. MIT Press. Schmitt, Viola (2019): Cross-categorial plurality and plural composition. Semantics and Pragmatics 12(17) Talk slides |
| December 2, 2020 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Mia Windhearn Heinrich Heine University Exclusive Quantification without Focus Alternatives This talk will discuss quantification by exclusive operators and the availability of semantic alternatives both with and without an overt focus structure. Exclusive operators like only have traditionally been assumed 1) to associate with focus, and 2) to quantify over propositions (Rooth 1992). I will argue that most focus-sensitive operators should more accurately be characterized as alternative-sensitive. Focus sensitivity is thus a sub-type of a broader phenomenon of alternative semantics, even in the context of exclusive operators. Much of the data comes from English just and its wide range of uses as canonical and non-canonical exclusive, some of which are exemplified below. As indicated, only a subset of these uses make use of a prosodic focus structure. (1) The lamp just broke! (All by itself!) (Unexplanatory) (2) Bob is just a [philosopher]F (not a linguist). (Focus-Evaluative) (3) Carl just has [two]F degrees. (Focus-Entailment) (4) Ana has just gone to get her car. (Temporal) (5) Thereճ a spider just above your head. (Spacial) (6) That spider is just gigantic! (Emphatic) In addition to English just, I will discuss some data from Serbian and Chկl that further support the separation of alternative set generation and focus. I will provide a unified semantics for exclusivity that takes exclusive operators to take arguments of variable type. In doing so, following Zimmermann 2017, I propose that exclusives should not be taken to quantify over propositions. I argue instead that exclusive operators take a structured proposition including the Ԭocusՠof alternation and the background question as their argument. Focus is one mechanism for providing such a locus, but as I hope to show, by no means the only strategy available to exclusive operators. Talk slides |
| November 18, 2020 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Todd Snider Heinrich Heine University Constraints on Propositional Anaphora Talk slides Anaphors are words whose reference is determined on the basis of the interpretation of some other word or phrase (its antecedent). For example, pronouns are anaphors whose antecedents denote individuals, as in (1). (1) Nancy has a car. She has owned it for five years. This talk will focus on anaphors which refer to propositions, as in (2). (2) Nancy has a car. She told me that. In particular, this talk will discuss some of the constraints on propositional anaphora, exploring when propositional anaphora is licit (and when it is not). The first part of the talk deals with pragmatic discourse-level constraints, in particular at-issueness. There are competing notions of at-issueness, but many are discussed in the literature—explicitly or implicitly—as having consequences for a proposition’s availability for anaphora in non-trivial ways (e.g., AnderBois et al. 2013). I argue against this tight linking, and demonstrate that a proposition’s at-issue status in a discourse (at least as defined by Simons et al. 2010) is neither necessary nor sufficient to determine its availability for anaphora. Thus, propositional anaphora is not constrained by atissueness, at least under one prominent definition thereof. The second part of the talk moves from the discourse to the sentence, to see if there are syntactic constraints on propositional anaphora. In the tradition of Karttunen’s (1969) examination of which NPs make an individual available for anaphoric reference, I present (highlights from) a comprehensive examination of which structures make propositions available for anaphora. I present some surprising results which cut across traditional syntactic classifications, including data on small clause and raising/control/ECM verb constructions. This data leads me to propose a new generalization for when propositions are available for anaphoric reference. I then compare the observed behavior of propositional anaphora to that of individual anaphora. |
| November 4, 2020 12:30-14:00, online via Webex https://hhu.webex.com/meet/tsnider Peter Sutton Heinrich Heine University A non-identity does not imply a plurality: The Problem of the Many and the semantics of numeral constructions Talk slides The Problem of the Many (Geach, 1962; Lewis, 1993; Unger, 1980) is a paradox that arises in relation to counting and/or quantifying over entities in the denotations of count nouns. For example, there are contexts/situations in which it is true that there is one cloud in the sky, however, when it comes to the extension of the cloud in question, there are multiple, non-identical sums of water droplets, each of which is an equally good candidate to be the extension of that cloud. Therefore, either each of these sums of droplets is a cloud and we have many clouds or no sum of droplets is a cloud and so we have no clouds. Either way, paradoxically, we do not have one cloud, contrary to our starting assumption. Perhaps surprisingly, the Problem of the Many is not usually discussed in relation to the semantics of numeral constructions such as one/two/three clouds. In this talk, I argue that the Problem of the Many forces us adopt an analysis of count nouns and numeral phrases based on a logically weaker mereological property than is commonly assumed. However, once we do so, it becomes clear what the source of the paradox is and how to avoid it: when it comes to counting with natural language predicates, it is not the case that a non-identity implies a plurality. In other words, we have good reason to think that if a and b are in the extension of P and ab, it is not always the case that the sum of a and b counts as two Ps. |
| January 15, 2020 Aviv Schoenfeld Tel Aviv University The subkind reading of nouns and the hierarchical structure of kinds The research question is: which nouns in English can get a subkind reading? The key data is in (1-4). I will present experimental support to the contrast between wild animal – wildlife in (1)-(2), which parallels that of vehicle-transport, treated in Sutton & Filip (2018: ex.5): (1) I wonder which wild animal is the most common. (2) #I wonder which wildlife is the most common. (3) I wonder which cheese is the most common. (4) I wonder which food is the most common. From Grimm & Levin (2017) and Sutton & Filip (2018), I derive the hypothesis that a noun can get a subkind reading if and only if it heads a certain hierarchical structure. I formulate the relevant structure based on data like (5)-(7), regarding which kinds can be in the denotation of nouns with a subkind reading: (5) Mutts are the most popular dog in America. (Mutts are dogs that are unclassified for breed.) (6) Knives are a prominent weapon in ancient Indian history. (Not every knife is a weapon.) (7) Grass is the most valuable plant on the planet. (Not everything that counts as grass is a plant organism.) Under my analysis, for count nouns like animal in (1), it does not matter whether the subkinds are named by count or mass nouns, but it does matter for mass nouns like those in (2)-(4). I will show that cheese and food head the relevant structure, because there are enough subkinds that are named by mass nouns (e.g. cheddar and mozzarella, meat and dairy). The same does not hold for wildlife in (2), e.g. English has no mass noun counterpart of tiger. Thus, I provide an alternative analysis to those in Grimm & Levin (2017) and Sutton & Filip (2018) regarding why object mass nouns like furniture and wildlife lack a subkind reading. Grimm, Scott & Beth Levin. 2017. Artifact nouns: Reference and countability. Proceedings of the North East Linguistic Society (NELS) 47. 55-64. Sutton, Peter R. & Hana Filip. 2018. Restrictions on subkind coercion in superordinate object mass nouns. Proceedings of Sinn und Bedeutung 21. 1195-213. |
| December 4, 2019 Jeremy Kuhn Institut Jean Nicod, CNRS cole Normale Suprieure Gather/numerous as a mass/count opposition Predicates like gather and be numerous have both been described as ԣollective predicates,ՠsince they predicate something of a plurality. The two classes of predicates differ, however, with respect to plural quantifiers (e.g. all), which are grammatical with gather-type predicates but ungrammatical with numerous-type predicates. Here, I show that the gather/numerous opposition derives from mereological properties that are familiar from the domains of telicity and mass/count. I address problems of undergeneration and overgeneration with two technical innovations: first, I weaken the property of divisibility to Champollionճ (2015) property of stratified reference; second, I provide mechanisms to rule out accidental satisfaction of the logical property. From a broader perspective, we place collective predication in a larger context by building empirical connections to mass/count and collectivity across semantic domains. |
| July 3, 2019 Alexander Stewart Davies University of Tartu Contingent Content Parthood and the False Dichotomy between Speech Act Pluralism and Speech Act Monism In a string of publications dating back to 1997, Cappelen and Lepore have used the fact that there can be multiple true indirect speech reports, each with a different complement, but nonetheless, made of a single utterance of a sentence, to defend the view that when we utter a sentence, we donմ just say one thing (the content of the sentence uttered), we thereby say many things. In this paper, I argue that neither view is correct. I argue that Ҵhing that A saidӠis a homogeneous count noun phrase, and for that reason, just as whether there is one or three bouquets in my hand depends upon the context for counting, so too, how many things a person says in uttering a sentence in context, depends on the context for counting. In the course of the argument, I will need a notion of content parthood. I provide a novel definition of content parthood which makes content parthood contingent. Thus, the parts of what someone says depend upon the world in which she says them. Ignorance of the world in which one says something, results in ignorance of (the parts of) what one is saying. |
| June 26, 2019 Yasutada Sudo University College London The Plurality Inference as a Non-Propositional Quantity Implicature Plural nouns typically give rise to ‘plurality inferences’, e.g. “Andrew wrote papers” implies that Andrew wrote multiple papers, not just one. However, plurality inferences are not always present, e.g. “Andrew did not write papers” does not mean the same thing as “Andrew did not write multiple papers”. There are three types of approaches to the plurality inference: (i) the scalar implicature approach (Spector 2007, Zweig 2009, Ivlieva 2013, Mayr 2015), (ii) the ambiguity approach (Farkas & de Swart 2010, Grimm 2013, Mart 2018), and (iii) the antipresupposition approach (Sauerland 2003, Sauerland et al. 2005). In this talk, I will propose a new scalar implicature account. The idea of the scalar implicature account is that the plural is semantically number-neutral, and the plurality inference arises as a scalar implicature in competition with the singular. Previous studies pointed out that the scalar implicature computation is not so straightforward, given that pairs like “Andrew wrote papers” and “Andrew wrote a paper” will be truth-conditionally identical, if the plural is semantically number-neutral. Different versions of the scalar implicature account make use of different truth-conditional asymmetries, e.g. non-global levels of meaning, strengthened meaning for the singular alternative, etc. I propose instead that the plurality inference can be derived as a global-level quantity implicature based on a non-truth-conditional aspect of the meaning. Specifically, the singular and plural sentences differ in anaphoric possibilities: the plural sentence introduces a discourse referent that ranges over singular or plural entities, while the singular sentence introduces a discourse referent that only ranges over singular entities. Based on this asymmetry, a quantity implicature is derived that the discourse referent is meant to only range over plural entities. This analysis requires no additional mechanism beyond what is usually assumed (embedded scalar implicature or higher-order implicature), unlike the previous scalar implicature accounts. I will formalize this idea in Update Semantics, and demonstrate that it makes correct predictions about negative sentences and quantified sentences. |
| June 12, 2019 Nina Haslinger University of Gttingen Viola Schmitt University of Vienna Cumulative readings of quantifiers: A plural projection approach We present an analysis of cumulative readings of quantifiers in German within the Plural Projection framework of Schmitt (2017). In this theory, the property of semantic plurality is generalized across categories: In addition to pluralities of individuals, there are pluralities of functions, propositions etc. Further, when a constituent X dominates a semantically plural expression Y, X counts as a plural expression as well and its denotation inherits the part structure associated with Y via a special composition rule. Cumulative readings are the result of repeatedly applying this composition rule. Unlike earlier approaches to cumulativity, this analysis is surface-compositional and does not make crucial use of event semantics. We will apply this framework to two types of examples. The first type involves universal quantifiers that have cumulative readings wrt. syntactically higher plurals, but not wrt. syntactically lower plurals. For instance, the ‘every’ DP in (1-a) has a cumulative reading, while the one in (1-b) does not (Schein 1993, Kratzer 2000 a.o.). (1) a. Two girls fed every dog in this town. b. Every girl in this town fed two dogs. In German, this class includes DPs with the determiner ‘jed-‘ (‘every’) as well as distributive conjunctions with ‘sowohl … als auch’ (‘as well as’). The second type of examples involves cumulative readings of modified numerals like ‘genau n Leute’ (‘exactly n people’) or ‘gerade mal n Leute’ (‘no more than n people’). As noted by Buccola & Spector (2016), the upper bounds introduced by numeral modifiers seem to disappear in some cumulative sentences: (2) Gerade mal fnf Leute haben mehr als 20 Liter Bier getrunken. no.more.than five people have more than 20 liters beer drunk ‘No more than five people drank more than 20 liters of beer.’ (2) can be true in scenarios where more than five people drank beer. We argue that such non-upper-bounded readings are only available if the modified-numeral DP cumulates with another plural in its scope. If so, cumulative readings of universal quantifiers and cumulative readings of modified numerals are both subject to semantic asymmetries conditioned by scope. So both classes of quantifiers provide evidence that cumulativity does not entail scopelessness. In the Plural Projection framework, interactions between cumulativity and scope are unsurprising since cumulative sentences do not involve a special LF syntax (as proposed by Beck & Sauerland 2000). The theory allows us to give a natural account of the asymmetry in (1) and can be extended to cover examples like (2) by combining it with a two-dimensional semantics for modified numerals. We will also show how this framework accounts for the `mixed’ cumulative/distributive readings discussed by Schein (1993). Finally, we will provide an independent argument against analyses based on syntactically derived cumulative predicates. The argument involves cumulative readings of sentences in which plural quantifiers occur within a conjunction. References Beck, Sigrid and Uli Sauerland (2000): Cumulation is needed: a reply to Winter (2000). Natural Language Semantics 8(4): 349-371. Buccola, Brian and Benjamin Spector (2016): Modified numerals and maximality. Linguistics and Philosophy 39: 151-199. Kratzer, Angelika (2000): The event argument and the semantics of verbs. Manuscript, UMass Amherst. Schein, Barry (1993): Plurals and Events. Cambridge/MA: MIT Press. Schmitt, Viola (2017): Pluralities across categories and plural projection. Manuscript, University of Vienna. |
| April 17, 2019 Massimo Poesio Queen Mary University of London Disagreements in anaphoric interpretation The assumption that natural language expressions have a single, discrete and clearly identifiable meaning in a given context, successfully challenged in lexical semantics by the rise of distributional models, nevertheless still underlies much work in computational linguistics, including work based on distributed representations. In this talk I will first of all present the evidence that convinced us that the assumption that a single interpretation can always be assigned to anaphoric expression is no more than a convenient idealization. I will then discuss recent work on the DALI project that aims to develop a new model of interpretation that abandons this assumption for the case of anaphoric interpretation / coreference. I will present the recently released Phrase Detectives 2.1 corpus, containing around 2 million crowdsourced judgments for more than 100,000 markables, an average of 20 judgments per markable; the Mention Pair Annotation (MPA) Bayesian inference model developed to aggregate these judgments; and the results of a preliminary analysis of disagreements in the corpus suggesting that between 10% and 30% of markables in the corpus appear to be genuinely ambiguous. |
| April 10, 2019 Ai Taniguchi Carleton University Non-redundancy, social meaning, and role language In this talk I use tools from formal semantics and pragmatics to analyze social meaning, and argue that social meaning is an imposition on the input context of an utterance. Particularly, I explain why the repeated usage of a sociolinguistic variant in discourse is not redundant. I then extend this analysis to an understudied phenomenon in Japanese called role language. Social meaning indexes a speakerճ identity and stances, including but not limited to geographical origin, age, gender, sexual orientation and even friendliness (cf., Eckert 2000; 2008; among others). For example, the monophthongized version of /paI/ realized as [pa:] and /ɹaIli/ as [ɹa:li] in (1) both index southernness. I call expressions like [pa:]/ [ɹa:li] socially indexing expressions (SIEճ). (1) Riley [ɹa:li] bought a pie [pa:]. Ԓiley bought a pieՠ+ Ԕhe speaker is from the South.ռo:p> Some scholars wonder if social meaning is a conventional implicature (CI) (Levinson 1979; Smith et al. 2010): a secondary entailment that comments on the primary (at-issue) entailment (Potts 2005). While I agree with Smith et al. (2010) that social meaning is a secondary entailment, I argue that itճ not a CI. Formally, I will analyze the social meaning of SIEճ as the intersection of its indexical field (the set of properties associated with the variant (Eckert 2008)) and what I call the indexical set, which is the set of all possible possible personas for the speaker, according to the hearer. This has the effect of making social meaning hearer-dependent: no matter what persona you want to portray to the world, what persona you actually end up conveying depends on how the hearer perceives you (Burnett 2019). Role language refers to a set of linguistic expressions that depict the speakerճ ңharacter type,Ӡlargely seen in Japanese fiction (Teshigawara & Kinsui 2011; Kinsui 2003). For example, -nya in (2) indicates that the speaker is a cat character. (2) nyaa-tachi-no mokuteki-wa mezurashii pokemon … sore-o wasureru-nya! me.nya-PL-GEN objective-top rare Pokemon that-ACC forget-NEG.NYA ԏur objective is (to catch) rare Pok ́emon. . . donմ forget that!ՠ+ Ԕhe speaker is a catռo:p> I argue in this talk that role language is really a kind of social meaning, contrary to Teshigawara & Kinsui (2011)ճ suggestion that social meaning does not include traits like speaker species. Something like -nya is analyzable as a SIE with a fairly specific indexical field: {λxλw.catw(x)}. It just means that the speaker is speaking a (faux) cat dialect. This work contributes to recent efforts by scholars (e.g., Beltrama & Staum Casasanto 2017; Burnett 2019; Smith et al. 2010) to bridge the gap between the sociolinguistic literature and the formal semantics/pragmatics literature. From this we gain insights into how different layers of meanings interact in language and manifest in identity construction. |
| April 3, 2019 Henk Zeevat SFB991, HHU Dsseldorf and ILLC, Amsterdam University Semantics made easy(?): Non-lexical frame semantics Standard logical representation of NL meaning starts from logic as given by the logical tradition (Boole, Frege, Kripke) and takes the basic organisation of standard logical systems (variables, boolean operators, quantifiers) as given. It is very questionable though that any of the logical operators in that tradition can be found in natural languages as constants. Non-lexical frame semantics (NLFS) is an attempt to come up with a different style of logical representation that is closer to the syntactic organisation of NLs than, e.g., DRT and that abolishes variables and logical operators by developing a new account of scope phenomena. NLFS is a geometrical model based on the graphs of frame semantics or dependency grammar (this taking on board both the Frame Hypothesis and the essential identity of syntactic representation with logical representation) and adding two new arrows: domina arrows to indicate that an element depends on another element and anaphora arrows that mark the relation between an anaphoric element and its antecedent. Negation is treated by introducing the notion of a negative predicate that excludes values for its non-dependent arguments. NL conjunction is treated by a combinatorial extension of the standard representations. The system is intended as a linguistic account of logical representation. It should be better in dealing with semantic universals and in capturing linguistic generalisations, e.g., between quantification, negation and conditionals, as well as aiming for simplicity. Since it also covers new ground, it constitutes a challenge to other proposals for logical representation of NL meaning. |
| January 9, 2019 Scott Grimm University of Rochester Plurality and Referentiality Certain uses of the plural form of indefinite nouns, such as in (1), have long been puzzling because they appear to include reference to singular entities. (1) a. Q: Do you have children? A: Yes, I have one. b. Ed didn’t see dogs. (False if Ed saw one dog) Based on such examples, many researchers have argued that plural indefinite nouns are actually “inclusive”, designating ‘one or more’ (including singular entities), rather than “exclusive”, designating ‘more than one’ (excluding singular entities). Sauerland et al. (2005), among others, relate the inclusive reading to downward-entailing environments (such as negation or the antecedent of conditionals), claiming number ҭarking on indefinites in a downward entailing environment does not affect truth conditionsӮ This talk presents experimental and empirical evidence that the inclusive readings (i) cannot be causally related to downward-entailing environments, as this association both under- and over-generates and (ii) cannot be associated with indefinites in general, but only with non-specific/non-referential indefinites. Instead, the experimental results indicate that the key factor is whether the indefinite can be construed as non-referential/generic, which favors inclusive readings, or referential, which resists them. The contexts which permit inclusive readings, e.g. negation and interrogatives, are shown to be just those which may in general block referential commitment. I argue this lack of referential commitment promotes inclusive plural readings. I then discuss the typological prediction of this account, namely that inclusive plural readings should only occur in languages for which the bare plural has non-referential/generic uses, a claim which receives support from languages such as Armenian. |
| December 12, 2018 Jakub Kozakoszczak University of Warsaw and Heinrich Heine University, Dsseldorf Natural Negative Evidence in Formal Semantics The common type of negative evidence employed in formal semantics is the judgment-based ҳomewhat elusive notion of semantic deviance –- a notion that the field of formal semantics acknowledges but might lack the right tools to modelӠ(Vecchi et al. 2011). қW]e still do not have a precise linguistic account of what it means for a linguistic expression to be ҮonsensicalӠor semantically deviant, nor a clear relation between this notion and that of being unattested in a corpus: Semantic deviance remains a difficult and understudied phenomenonӠ(Vecchi et al. 2016). In this talk, I will present (i) an analytic procedure of translating semantically unacceptable expressions to factorized comprehension sets (Katz 1964) via intended meaning and semantic hypothesis, and (ii) a statistical test to tell whether the expression has zero probability of occurring in text corpora given the meaning and the hypothesis. I will demonstrate this treatment on starred expressions from the existing literature. The resulting new type of evidence is the empirical probability that a given formalized meaning is impossible to occur. I will discuss its limitations and argue for its potential to address four flaws of the traditional judgment-based evidence: 1. Volatility. The differences in the judged acceptability among respondents can originate from different meanings ascribed to the expression. 2. Causal obliqueness. The psychological relation of unacceptability is equivocal in that it can be caused by different flaws, suit different explanations, and support unrelated hypotheses. 3. Unobservability. Judgments are produced by respondents through introspection whereas text corpora are intersubjectively observable. 4. Theory-loadedness. Judgment reports are responses to designed questions whereas language behavior documented in corpora occurs naturally without interaction with the researcher. |
| December 5, 2018 Federico Silvagni Complutense University of Madrid States, events and copulas: a Romance perspective This talk focuses on the aspectual distinction between States and Events and its manifestations in Romance languages, with particular attention to Spanish. The main goal of this study is to establish a satisfactory limit between these two aspectual classes, by answering two main questions: (i) what is an Event (and, as a consequence, a State)? (ii) How is this distinction encoded in grammar? The need for this study derives from the puzzling (and long-standing) observation that [dynamism], or any other equivalent criterion (Vendler 1957; Kenny 1963; Comrie 1976; a.o.), is not a relevant primitive of eventivity, since there are predicates that behave as Events despite the fact that they are non-dynamic, i.e. static (Dowty 1979; Maienborn 2005, 2007, 2011; a.o.). I observe that, once we abandon [dynamism], we reach a State/Event distribution that is equivalent to the Individual-Level/Stage-Level distinction (Milsark 1974; Carlson 1977), whose understanding, in fact, represents another major unknown for the research on inner aspect (see Fernald 2000; Fbregas 2012). I put forward the hypothesis that the State/Event and the Individual/Stage distinctions are one and the same thing (see also Hoekstra 1992), and that the difference between the two classes rests merely on the lack or the presence of inner aspect. Moreover, based on the concept of ԥventՠadopted in modern (post-Einsteinian/Minkowskian) physics and philosophy, where reality is taken as a 4D continuum (3 Space + 1 Time dimensions), I propose that the aspectual primitive of eventivity is a ԓpacetime pointլ which I label as [Stage] (also Silvagni 2015, 2016, 2017). In order to find out how such a [Stage] primitive (and, thus, the State / Event distinction) is encoded in grammar, I focus on non-verbal predication in Romance languages; more in particular, I study the Spanish copular alternation (ser/estar) and its relation to mono-copular languages, such as French and Italian. I empirically show that [Stage] is a formal feature (in the sense of Zeijlstra 2008, 2014), which is encoded in Event predicates (that is, SLPs) as an uninterpretable instance [uS], and that Eventive (or SL) structures are derived by an agreement operation between a predicate and an Asp head endowed with an [iS] feature, which is realized as estar in the case of Spanish. The study provides a more accurate understanding of the State/Event distinction and, at the same time, the Individual/Stage contrast. Additionally, it constitutes a detailed analysis of non-verbal predication and copular alternation in Romance, which can also offer a comprehensive explanation of typical controversial phenomena, such as coercion and the IL/SL distinction in those languages that lack a SL-copula. |
| November 21, 2018 Radek Šimk Humboldt-Universitt zu Berlin Interpreting Slavic bare NPs: Empirical reflections on the null hypothesis The literature on bare NPs in articleless languages has been dominated by the neo-Carlsonian view, according to which bare NPs denote properties and can be type-shifted to other meaning types, such as particulars (iota), kinds (cap), or quantifiers (ex), whereby these type-shifts are restricted in various ways (Chierchia 1998, Krifka 2003, Dayal 2004, Geist 2010, a.o.). According to some versions of this view, bare singulars end up being interpreted on a par with definite descriptions, either essentially always (Dayal 2004) or always when they are topical (Geist 2010). In my talk, I will offer an alternative, inspired by Heimճ work (Heim 1982, Heim 2011), according to which bare NPs in articleless languages, including bare singulars, are completely underspecified with respect to (in)definiteness. The idea is that bare NPs denote restricted variables (or functions from situations to such variables, being of type <s,e>), which either receive a value contextually or are bound by intra-sentential quantifiers (Heim 1982). Consequently, inferences typically associated with definite descriptions – particularly uniqueness and/or maximality – never really arise for bare NPs (contra the neo-Carlsonian view of ҤefiniteӠbare NPs). This return to what could be considered the null hypothesis – namely that definiteness is not even a relevant category in languages without articles – will be supported by corpus and experimental evidence. Novel analyses based on Czech corpus data (from Šimk & Burianov 2018) show that bare NPs are interpreted as indefinite more often than some theories would expect. Also, they exhibit no bias of singulars to be interpreted as definites (Dayal 2004) – the proportion of definite and indefinite bare singulars vs. plurals is identical. Furthermore, the majority of indefinite interpretations of Czech bare NPs can be traced to particular expressions in the clause (e.g. presentational verbs, negation, adverbs of quantification, etc.) – suggesting that indefinite interpretations indeed arise from clause-internal quantificational closure of the bare NP variable, contributed by a suitable object language expression. Indefinite bare singulars are thus no longer quirks requiring special treatment (Dayal 2004), but fall out naturally from the analysis. Another portion of evidence comes from experiments on the interpretation of Russian bare NPs, as compared to German (in)definite NPs (Šimk & Demian, submitted). These experiments show a weak but consistent effect of definiteness on uniqueness (sg) and maximality (pl) in German. Russian, on the other hand, exhibits no such consistent effect – Russian bare NPs are interpreted as indefinite (or non-definite) even if such an interpretation is primed by information structure (word order, prosody; cf. Geist 2010 and many others), or grammatical number (Dayal 2004), making the results fully consistent with the null hypothesis. |
| November 7, 2018 Halima Husič Ruhr-Universitt Bochum Towards a Formal Account of the Countability of Qualities and Events While abstract objects are a well-discussed topic in philosophy, in linguistics this class of nouns has occasionally been investigated, usually as part of a bigger linguistic issue, as e.g. derived nominals, countability or polysemy. In standard analyses of the semantics of count and mass nouns ԡbstract nounsՠare usually left out. However, ԡbstract nounsՠare frequent in language and they can also be count or mass: (1) Have you noticed how much beauty goes into dying? (2) One of the many virtues of pumpkins is the ability to combine equally well with sugar and spices or salt and cheeses, making them perfect for pies, cookies and cakes, as well as for soups, side dishes and stews. In my talk I will discuss the polysemy of nouns that denote abstract entities and tackle the question how countability is defined in such nouns. I will present an empirical investigation of such nouns which includes (i) an annotation of lexical properties that are responsible for a count and mass interpretations, (ii) a specification of dependent senses of nouns and how they are derived and (iii) a corpus study. Based on this research I will propose an analysis to account for count and mass uses of ԡbstract nounsծ |
| October 24, 2018 David Nicolas CNRS, Institut Jean Nicod (joint work with Salvatore Florio, University of Birmingham) Plurals and mereology A prominent tradition in linguistic semantics analyzes plurals by appealing to mereology (e.g. Massey 1976,Link 1983, Link 1998, Gillon 1992, Champollion 2016, Champollion 2017). The mereological approach to the semantics of plurals has attracted heavy criticisms, especially within philosophical logic where many authors subscribe to an alternative framework based on plural logic (e.g. Boolos 1984, Yi 1999, Rayo 2002, Rayo 2006, McKay 2006, Oliver & Smiley 2013). Some of the criticisms target a broader class of ‘singularist’ semantics which interpret plural expressions in terms of singular ones. The mereological approach is the most popular, and perhaps the most plausible, of those analyses. These criticisms have been very influential in philosophy. What has been overlooked is that, once the mereological approach is properly understood, linguists have the basic tools to respond. Our aim is to clarify and develop these responses systematically, bringing together the linguistic and philosophical literature. In the first part of the talk, we offer a precise articulation of the mereological approach and of the semantic background in which the debate can be meaningfully reconstructed. We focus on the best-known implementation of the approach, namely that of Godehard Link, but we comment on the relevance of our discussion for alternative implementations. In the second part, we deal with the criticisms and assess their logical, linguistic, and philosophical significance. Our conclusion is that the mereological approach fares no worse than plural logic and that it remains a viable and well-motivated framework for the analysis of plurals. |
| October 17, 2108 Carol-Rose Little Cornell University Numerals and quantification in Chկl In this talk, I investigate a set of underdescribed numeral constructions in Chկl that I call possessed numerals. Unlike the bare numerals in (1), these possessed numerals in (2) occur with possessive prefix i- and relational suffix -el, allowing both cardinal (2a) and ordinal (2b) interpretations. (1) chaխtyikil wiik two-CL man Դwo menռo:p> (2) Possessed numerals a. tyi i-chaխtyikl-el wiik TYI A3-two-CL-RS man Դhe two menռo:p> b. i-chaխtyikl-el=b wiik A3-two-CL-RS=REL man Դhe second manռo:p> The cardinal interpretation in (2a) contains tyi in addition to the possessed numeral and the ordinal interpretation in (2b) contains a relative clause marker =b. Building off the semantics of bare numerals in Chկl (Bale and Coon, 2014), I propose a compositional semantics for these possessed numerals. I argue that tyi in (2a) introduces a uniqueness presupposition, supported by data from the ordinal construction and the possessed numeral ԯneլ which lack tyi. The morpheme tyi shows up in other quantificational phrases to derive strong quantifiers from weak quantifiers. For instance, I present data from other quantifiers like pejtyel and its possessed form tyi i-pejtyel-el ԡllՠthat have similar entailment differences as (2a) and (1) do. All data comes from my fieldwork in Chiapas, Mexico. |
| July 12, 2018 Lucas Champollion New York University Two switches in the theory of counterfactuals: A study of truth conditionality and minimal change I report on a comprehension experiment on counterfactual conditionals based on a context involving two switches (joint work with Ivano Ciardelli and Linmin Zhang). We found that the truth-conditionally equivalent clauses (i) switch A or switch B is down and (ii) switch A and switch B are not both up make different semantic contributions when embedded in counterfactual antecedents. Assuming compositionality, this contradicts the textbook view that meaning can be identified with truth conditions. This finding has a clear explanation in inquisitive semantics: truth-conditionally equivalent clauses may be associated with different propositional alternatives, each of which counts as a separate counterfactual assumption. Related results from the same experiment challenge the common Stalnaker-Lewis interpretation of counterfactuals as involving minimizing change with respect to the actual state of affairs. We propose to replace the idea of minimal change by a distinction between foreground and background for a given counterfactual assumption: the background is held fixed in the counterfactual situation, while the foreground can be varied without any minimality constraint. (This talk presents work reported in a paper recently appeared in the journal Linguistics and Philosophy, available at http://ling.auf.net/lingbuzz/003200.) |
| July 4, 2018 Eytan Zweig University of York Semantic and Pragmatic Effects on Bare Plurals in Questions It is well established in the literature (see Sauerland et al. 2005 and Zweig 2009 among others) that in the context of a polar question, bare plurals and singular indefinites do not seem to be distinct; i.e., (1a) and (1b) will receive identical answers, ҙesӠfor at least one apple and ҎoӠonly if there are no apples whatsoever: 1a. Do we have apples in our pantry? 1b. Do we have an apple in our pantry? What is less frequently discussed in the literature, however, is that this similarity is modulated by a variety of pragmatic considerations, relating to both the context in which the question was asked and the possible referents for the noun phrase in question. In this talk, I will explore some of these effects, presenting novel data from a visual world experiment and discussing the theoretical implications of its results in the greater context of plural semantics. |
| June 27, 2018 Luka Szucsich Humboldt‐University Berlin Obviation in North Slavic languages In many languages, including Romance and some Slavic languages, subjunctive clauses selected by volitional verbs allow for syntactic phenomena to occur in a less local domain than it is the case with other types of subordinate clauses (especially indicative clauses). One of these phenomena is so‐called ‛obviationլ i.e. obligatory disjoint reference of subjects in the matrix clause and subjects in subjunctive clauses, cf. (1) for Polish. (1) Jareki chc‐e, żeby pro*i/j śpiewa‐ł. Polish JarekM:SG:N wantPRS:3:SG thatSUBJ pro singLPT‐M:SG ‛Jarek wants him to sing.ռo:p> Obviation is often attributed to ‛domain extensionԠcollapsing the tense domains of the matrix and the embedded clause. The nature of domain extension, however, is rarely explicitly analyzed. I will present an analysis of obviation effects which relies on cross‐clausal feature dependencies. I assume that (potentially cross‐clausal) occurrences of features may constitute the relevant domains for construal processes. Selectional properties of volitional verbs and the deficiency of the clausal complementԳ T‐feature provide the prerequisite for cross‐clausal feature sharing, i.e. the matrix and the complement clauseԳ T‐features form one feature chain making the respective instances featurally identical. I will show, however, that feature identity may be not only determined by φ‐features and referential indices, but also by features such as [agentivity], or features determining, whether a situation is subject to control by a participant (Szabolcsiճ RESP feature), or contrastive focus features. |
| June 13, 2018 Marcin Wągiel Masaryk University Brno Subatomic quantification in natural language In standard lattice‐theoretic approaches to natural language (e.g., Link 1983, Landman 2000, Champollion 2017) singularities and pluralities are presumed to involve distinct mereologies and it is commonly supposed that quantificational expressions do not access subatomic part‐whole relations. In this talk, I explore three hypotheses regarding natural language semantics: (i) natural language is sensitive to topological relations holding between parts of singularities (cf. Grimm 2012), (ii) there are general counting rules that presuppose such relations, and (iii) quantification over parts is subject to identical restrictions as quantification over wholes. I propose that Դo be countable’ means Դo be an integrated part of an integrated whole’ where art’ employs the reflexive mereological relation, and thus covers both proper and improper parts. In other words, only an entity conceptualized as a whole that comes in one piece can be put in one‐to‐one correspondence with the natural numbers. Similar, a proper part of such an entity that is conceptualized as an integrated object is also countable. The evidence for the linguistic relevance of this claim comes from the interaction between cardinal numerals and partitives involving irregular plurals in Italian as well as distinct classes of Polish Ԩalf’, art’ and Էhole’ words and from the quantificational behavior of multipliers such as the English Ԥoubleծ |
| May 30, 2018 John Collins University of East Anglia The Syntax-Semantics Interface and Its Ontological Significance The orthodoxy in semantic theory is that semantics carries various presuppositions and entailments about ontology – the world. I shall suggest that all the virtues of the orthodoxy may be retained while relinquishing the ontological load. Semantics, on such a view, constitutes constraints upon what can be said rather than determinants of how the world is. |
| May 2, 2018 Jon Ander Mendia Heinrich Heine University, Dsseldorf Epistemic Numbers Epistemic Numbers (ENs) are complex cardinal numerals formed by combining a number with a non‐numerical expression, typically an indefinite. Unlike other cardinal numbers, ENs (i) denote a range of possible values, and (ii) convey an epistemic effect of uncertainty or vagueness about the exact number. For instance, a sentence like twenty‐some people came is an acceptable answer to a question like How many people came?, and it can denote integers ranging from 21 to 29, conveying that the speaker is unsure about the exact number. I focus on three main cross‐linguistic facts: (i) some languages allow ENs with a variety of quantifiers, not just indefinites akin to some or wh‐indefinites; (ii) in some languages ENs can either precede or follow the numeral; and (iii) the interpretation of ENs depends on the numerical base of the language, and cannot combine with just any numeral. Giving prominence to this cross‐linguistic variability, in this talk, I provide an initial semantic analysis flexible enough to account for a variety of languages, and a pragmatic account of the epistemic effect. |
| April 25, 2018 Henk Zeevat University of Amsterdam & Heinrich Heine University, Dsseldorf Quantification in Frames In earlier work, I discussed the Stat’imc’ets quantifier system in which no generalised quantifier can take scope over another and which therefore only has cumulative readings for QNP V QNP sentences. This gives an argument for frames as the basis for NL semantics rather than (extensions of) FOL, since, in the latter systems, quantifiers are predicted to have scope over each other by default and cumulative readings that have to be painfully reconstructed. The Stat’imc’ets data seem to show that rather than scope taking readings, cumulative readings are the typologically unmarked case for QNP V QNP sentences. Stat’imc’ets has however a category of formally distinct dependent NPs which together with possessive NPs allow for readings dependent on quantifiers. These readings need to be accounted for, just like the possibility of the Stat’imc’ets system leading to a system like English in which all quantifiers can be dependent. The talk will present a new treatment of this using the attribute domina which links dependent NPs to their dominant NP or operator (this is just the easiest way of dealing with quantifier scope in dependency grammar). Dominated NPs will end up having functional denotations in an extension of a standard satisfaction semantics for frames and will so also allow an explanation of the paycheck sentences. Also, the Barry Schein mixed cases can be adequately treated (cumulative but the tricks may be dependent on the basketball players): Every basketball player learnt two new tricks from these videos. Viola Schmitt’s pioneering work on conjunction and plurals can be constructed as an argument against my frame analysis of cumulative quantification, showing that it is much too simple and needs extension to most other categories. This is a highly non‐trivial business in frame semantics. An example is the cumulative reading (only Bill sang, only John danced) of: John and Bill sang and danced. A new generalisation is presented that overcomes the objection and that also deals with collective readings. The Stat’imc’ets quantifier system is a special case of the generalisation. A full treatment of conjunction in frames is however definitely an open problem. |
| April 18, 2018 Daniel Gutzmann University of Cologne I lost my damn watch! Expressive adjectives between syntax and semantics Expressive adjectives (EAs) received a lot of attention in the semantic literature, which however treated their syntax mostly just like that of ordinary descriptive adjective. However, as I show in this talk, EAs differ in many and sometime puzzling aspects from descriptive ones. They cannot be modified or carry degree morphology, for instance. Most crucially, they exhibit interesting compositional behavior since they can be interpreted in a position different than they appear. Arguing against a purely pragmatic approach, both experimentally and conceptually, I propose a syntactic solution for the constraints in the semantic interpretation of EAs that builds on the assumption that expressivity is a syntactic feature, on par with features like tense or number. I develop an (upwards) agreement‐based analysis according to which the expressive content of an EAs is introduced by an uninterpretable expressivity feature which agrees with a corresponding interpretable feature at some higher node. This not only gets the restrictions about where EAs can be interpreted right but can also derive a lot of their special behavior and leads to direct mapping between syntactic structure and semantic interpretation. |
| January 24, 2018 Louise McNally Universitat Pompeu Fabra, Barcelona ‘Figurative’ uses of verbs and grammar (joint work with Alexandra Spalek) The goal of this talk, based on work in progress with Alexandra Spalek, is to draw attention to the contrastive study of ‘figurative’ uses of verbs as a way of gaining a new perspective on the challenges that must be met when addressing the interaction between conceptual content associated with words and their grammatical properties. Specifically, I will present the results of a comparison of the ‘literal’ and ‘figurative’ uses of six English verbs and their Spanish counterparts. This comparison reveals that figurative uses are strongly constrained by grammatical properties of the literal uses, indicating that the two uses emanate from a single lexical entry for each verb. Crucially, we find striking contrasts between figurative uses in the two languages which correlate with independently identified contrasts in the grammars of their verbal systems. However, at the same time, we find striking similarities in the sorts of conceptual domains in which the figurative uses are attested for each pair, even when the grammatical details differ, pointing to important conceptual similarities between the English and Spanish counterparts. The data will therefore force us to ask how contrasting grammatical constraints are associated with similar conceptual content. I will offer some preliminary reflections on this question, particularly in relation to the debate over Levin & Rappaport Hovav’s (1991) Manner/Result Complementarity hypothesis. |
| January 17, 2018 Michael Daniel Linguistics Convergence Laboratory National Research University Higher School of Economics, Moscow ROOM: Building 25.13, Room U1.24 (PLEASE NOTE THE CHANGE OF VENUE) Associative plural as indexical category Animacy Hierarchy was introduced as a cross-grammatical factor which governs, in different languages, widespread splits of nominal categories, including case or number marking. Among other phenomena, Animacy Hierarchy has been invoked to describe the lexical distribution of associative plurals (APL). In this paper I argue that APL as an interpretation of nominal plurality is motivated not by the high position the respective noun holds on the Animacy Hierarchy but is a combined effect of coercion by the unique reference inherent to proper names and associative links easily recoverable for human referents. Interpretation of APL relies upon activation of set relations its referent holds to other entities in the speech act or in the speech participantsՊ minds. APL is thus argued to be an indexical rather than a functional semantic category. — In the end, and if we have time, I will place APL in the context of other interpretative, or ‘vague’ categories, such as flexible relative clauses and genitives. |
| January 10, 2018 Curt Anderson Heinrich Heine University ғomeӠExclamatives In previous work (Anderson, to appear, Proceedings of Sinn und Bedeutung 21), I provide an analysis of what I call ҳomeӭexclamatives, an exclamative form in English using the determiner ҳomeӠalong with an intonational contour marking it as an exclamative (as in Ҩe is some lawyer!ө. These exclamatives express a positive or negative evaluation of how the referent of the ҳomeӭheaded DP exemplifies the property denoted by the NP. In this talk, I present preliminary experimental data examining the interaction of rising/falling intonation and speaker evaluation. From an online judgement study, we find that listeners can use intonation to make inferences regarding whether the speaker is making a positive or negative evaluation. Additionally, we find interactions between bias and the semantic class of the nominal complement of ҳomeӠ(whether the noun denotes a human or an instrument). I extend my prior analysis in order to account for some of these findings. |
| December 13, 2017 CANCELLED WILL BE RESCHEDULED John Collins University of East Anglia The Syntax-Semantics Interface and Its Ontological Significance The orthodoxy in semantic theory is that semantics carries various presuppositions and entailments about ontology ‐ the world. I shall suggest that all the virtues of the orthodoxy may be retained while relinquishing the ontological load. Semantics, on such a view, constitutes constraints upon what can be said rather than determinants of how the world is. |
| December 6, 2017 Keren Khrizman Heinrich Heine University Russian Diminutives as Counting and Measuring Operators: Implications for the Semantics of Counting, Measuring and the Mass/Count Distinction Diminutive nominal suffixes occur in many typologically diverse languages and have a varied interpretational pattern. Crosslinguistic studies identify at least three uses of diminutives: (i) expression of smallness; (ii) expression of individuation into small/minimal units accompanied by a grammatical shift from mass to count noun; (iii) emotional evaluation (Jurafsky 1996, Schneider 2003/13, Fortin 2011). Previous research has focused on the contrast between proper diminutives expressing small size/measure and emotive diminutives (e.g. Schneider 2003/13, Fortin 2011) and there has been little semantic study of the individuating function of diminutives. The results of a few studies suggest that (at least in some languages) there may be a categorial distinction between the two types of suffixes: proper diminutives are adjectival (Fortin 2011, Wiltschko & Steriopolo 2007) while individuating suffixes are either quantificational (Jurafsky 1996, Fortin 2011) or classifiers (Wiltschko 2006, Ott 2011). However, this has not been explored in detail. The present project aims to explore the syntax and semantics of ҰroperӠvs. individuating diminutives in Russian (as in stolik ԡ small tableՠand ris‐risinka Բice‐ a grain of riceթ, focusing on the following questions: (i) Why do diminutives individuate, i.e. why does smallness facilitate grammatical individuation? (ii) Under what grammatical, lexical and pragmatic conditions do they individuate? (iii) Is the semantic contrast reflected in a syntactic, categorial distinction, and if so, then why? I explore these issues in the context of recent work on counting, measuring and the mass/count distinction (Chierchia 1998/2010, Krifka 1989/95, Landman 2004/16, Rothstein 2009/10/17, Filip & Sutton 2016) and hypothesize that individuating and non‐individuating diminutives belong to the class of counting and measuring operators respectively. More specifically, proper diminutives are measure modifiers assigning measure characteristics to entities, whereas individuating suffixes are counting operators which allow parts of substances to be counted. I show that studying diminutives in the proposed framework will lead to a better understanding of diminutives, and, also, it will throw light on fundamental topics in semantics including measuring, countability and individuation. |
| November 29, 2017 Kurt Erbach, Peter R. Sutton, Hana Filip, and Kathrin Byrdeck Heinrich Heine University Object mass nouns in Japanese Classiޥr languages are commonly taken to have no grammaticized lexical mass/count distinction, but rather have this distinction encoded through the syntax and semantics of classiޥrs (e.g. Chierchia 1998, 2010; Muromatsu 2003; Rothstein 2010, 2017). We contest this claim by drawing on data from Japanese, a typical classiޥr language. We provide novel empirical evidence showing that Japanese quantiޥrs (e.g. nan-byaku-to-iu Ԩundreds ofթ can be used as tests for the mass/count status of Japanese nouns in the absence of classiޥrs. Such quantifiers can directly modify nouns like onna-no-hito (Էomanթ denoting inherently individuable entities, but not nouns denoting undifferentiated stuff like yuki (Գnowթ. Moreover, and more importantly, acceptability tests using these quantiޥrs lead to the identification of object mass nouns in Japanese, i.e. nouns that have inherently individuable entities in their denotation and yet are infelicitous in syntactic environments which are diagnostic of count nouns. This mismatch between notional and grammatical categories and the existence of object mass nouns in Japanese contradicts Chierchia’s (2010) prediction that such noun behavior should not exist in classifier languages. It follows that Japanese lexical nominal system is endowed with a grammatical mass/count distinction, which bears a certain resemblance to that which we ޮd in number marking languages (e.g. English), therefore casting doubt on the common view that classiޥr languages have no grammaticized lexical mass/count distinction. We propose a novel semantic analysis of Japanese lexical nouns and classiޥrs, based on Sutton & Filip (2016), a framework that unites notions of context in Rothstein (2010) and Landman (2011), and motivates the idea that counting contexts can remove overlap so that count nouns have disjoint counting bases while mass nouns do not. |
| November 22, 2017 Yasutada Sudo University College London The semantic role of classifiers in Japanese In obligatory classifier languages like Japanese, numerals cannot directly modify nouns without the help of a classifier. It is standardly considered that this is because nouns in obligatory classifier languages have uncountable denotations, unlike in non-classifier languages like English, and they need to be turned into countable denotations by classifiers before being able to be modified by numerals. Contrary to this, it is proposed that what makes Japanese an obligatory classifier language is not the semantics of nouns but the semantics of numerals. Specifically, evidence is presented that numerals in Japanese cannot function as predicates on their own, which is taken as suggesting that numerals in Japanese are exclusively used as singular terms. It is then proposed that the semantic function of classifiers is to turn such singular terms into modifiers/predicates. |
| November 15, 2017 Laura Kallmeyer and Rainer Osswald Heinrich Heine University Polysemy, Coercion, and Quantification In this work we model systematic polysemy and its interaction with quantification within a framework that combines Lexicalized Tree Adjoining Grammar (LTAG) with frame semantics and Hybrid Logic (HL). We first present the proposal from Babonnaud et al. (2016) according to which an inherently polysemous noun such as `book’, which provides referential access to both a physical and an informational object, is assumed to refer to entities of type physical‐object which have an attribute content whose value is of type information. The physical aspect and the informational aspect are then addressed in different ways in contexts such as `read the book’, `carry the book’, `master the book’ or `a heavy book on magic’. Treating books as physical information carriers in this way gives rise to the following ұuantification puzzleӠas noted for instance in (Asher & Pustejovsky, 2006). While ‘John carried off every book in the library’ poses no problem since the domain of quantification consists of physical entities, it not obvious how to cope with ‘John read every book in the library’, which is naturally interpreted as quantifying over all contents of the books in the library. Since the library may own more than one copy of a book, there is not necessarily a one‐to‐one correspondence between the physical books in the library and the book contents. It is not even necessary that John used a copy from the library at all. The solution we propose retains the idea that books are treated as physical information carriers, i.e., we quantify over these physical objects, while using only the information components as arguments of the predicate in the scope of the quantifier. This is possible due to the flexibility of the chosen syntax‐semantics architecture, in particular, underspecified use of Hybrid Logic and specific syntax‐semantics interface features. |
| November 8, 2017 Eleni Gregoromichelaki and Jon Ander Mendia Heinrich Heine University Continuation and discussion of their respective talks on October 18 and 25, 2017. |
| October 25, 2017 Eleni Gregoromichelaki Heinrich Heine University Ad-hoc grammatical categorisation in DS-TTR I will introduce a formalism, DS-TTR, motivated by the need to underpin dialogue modelling and explain the role of language in human interaction. In DS-TRR, the view of natural languages as codes mediating a mapping between ҥxpressionsӠand the world is abandoned to give way to a model where utterances are taken as joint actions aimed to locally and incrementally alter the affordances of the context. Such actions employ perceptual stimuli composed not only of ҷordsӠand ҳyntaxӠbut also elements like visual marks, gestures, sounds, etc. Any such stimuli can participate in the domain-general processes that constitute the ҧrammarӬ whose function is to incrementally predict the next action steps via the dynamic categorisation and integration of various perceptual inputs. One consequence of this view is that specifically linguistic syntactic categories, representations or constraints are eschewed and explained away as the effects of the dynamics of interactive processing. Given these assumptions, a challenge that arises is how to account for the reification of such processes as exemplified in apparent metarepresentational practices like quotation, reporting, citation etc. I will argue that even such phenomena can receive adequate and natural explanations as ҤemonstrationsӠthrough a grammar that allows for the ad-hoc creation of occasion-specific content through reflexive mechanisms. |
| October 18, 2017 Jon Ander Mendia Heinrich Heine University Some kind of relative clause Amount Relatives (ARs) differ from restrictive relative clauses in that they have an interpretation where what is referred to is not a particular object denoted by the head of therelative clause, but an amount of such objects (Carlson 1977, Heim 1987). For instance, the sentence it will take us the rest of our lives to drink the champagne they spilled that evening is ambiguous: on its more salient interpretation, the sentence claims that what would take us long to drink is a particular ҡmountӠof champagne, not any ҰarticularӠchampagne. Virtually all extant analyses of ARs agree on one thing: the that-clause denotes a set of degrees. This is usually implemented by covert movement of a degree operator and abstraction over a degree variable. In this talk I propose to reanalyze ARs. I suggest that the "amount" interpretation is a special case of ҫindӊ interpretation. First, I show that both "kind" and ҡmountӊ interpretations show the same distinguishing properties when we compare them to ordinary (intersective) relative clauses. Second, I show that there is no trace of the presence of degree quantification/abstraction in ARs, a fact that, all else equal, is not expected if Ars were degree expressions. Then I analyze "amount" interpretations as cases of ad hoc ҳubkindӠinterpretations: the that-clause contributes a way to determine a particular subkind of the kind-level object provided by the head of the relative clause (in the example above, an ad hoc subkind of champagne). Time permitting, I will discuss the fact that other languages may possess ARs as envisioned by traditional analyses -i.e. they must be treated as degree expressions- and consider the cross-linguistic ramifications of this state of affairs. |
| July 26, 2017 Shalom Lappin University of Gothenburg, King’s College London and Queen Mary University of London Deep Learning and Semantic Interpretation of Natural Language Classical approaches to formal and computational semantics assign values to the terminal elements of hierarchical syntactic structures and define combinatorial operations on the semantic representations of phrases to compute the values of sentences. While these approaches offer formally elegant models of interpretation, they have not produced wide coverage systems. They do not provide for semantic learning. They have also not succeeded in integrating lexical and compositional semantics in an interesting or computationally efficient way. Recent developments in image caption generation suggest an alternative approach, which can overcome these difficulties. This work formulates the problem of matching images with descriptions as a task in machine translation. Deep neural networks use an encoder to map regions of pixels in an image to vector representations of graphic features, and a decoder to align these features with the distributional vectors of lexical and phrasal items. This approach can be generalized to deep neural networks that identify correspondences between multi-modal data structures and sentences. To the extent that this research program is successful, it will satisfy the core objective of the classical formal semantic program. It will assign truth (fulfilment) conditions to the sentences of a language, where these conditions are specified in terms of multi-modal representations of situations (scenes) in the world. These correspondences are generated not by a recursive definition of a truth predicate in a formal semantic theory, but by an extended deep neural language model. |
| July 19, 2017 Anna Czypionka University of Konstanz Temporal implicatures in sentence comprehension: Evidence from acceptability ratings and self-paced reading times Implicatures are inferences that go beyond the literal semantic meaning of an utterance. In psycholinguistics, scalar implicatures are the most widely researched type of implicature. These implicatures are triggered by scalar expressions, like some (implying some, but not all), and have been investigated in a number of paradigms. We present data on a new type of implicature, namely, temporal implicatures. These are triggered by temporal expressions like this week (implying this week, but not next). When combined with predicates that ascribe a temporary property like be ill, this implicature makes sense: John is ill this week implies that John is ill only this week. However, a combination with predicates ascribing permanent properties like be tall is less felicitous: # John is tall this week implies that John’s size is not constant; this implicature clashes with speakers’ world knowledge about the size of grown-ups. A series of judgment studies employing the Literal Lucy paradigm show that just like scalar implicatures, temporal implicatures are psychologically real and are independent from the semantic meaning of the sentence. In addition, data from self-paced reading show that temporal implicatures are routinely drawn during language comprehension, and computed early. We argue that the infelicity of temporal modification with permanent predicates (# John is tall this week) is due to a pragmatic inference that conflicts with world-knowledge, rather than grammatical. |
| July 12, 2017 Agata Renans Ulster University Accounting for inferences of pluralized count and mass nouns : Evidence from Greek Across languages, plural marking on count nouns typically gives rise to multiplicity inferences , indicating that there is more than one entity in the denotation of the noun. Plural marking has also been observed to occur on mass nouns in Greek, giving rise to a parallel abundance inference, indicating that there is a large quantity of what is denoted by the noun. Kane et al. (2016) propose a unified implicature account of abundance and multiplicity inferences, which prima facie predicts a uniform pattern across inferences, and standard implicatures. We tested this prediction by comparing multiplicity inferences, abundances inferences, and standard implicatures in Greek-speaking children and adults. The results reflect an overall pattern of implicature calculation, supporting a unified implicature analysis across the inferences. |
| July 5, 2017 Kurt Erbach Heinrich Heine University Fighting for a share of the covers: Accounting for inaccessible readings of plural predicatesPlural predication presents a challenge that remains unsettled despite the numerous attempts to present a satisfying analysis (Moltman 2016, Farkas &de Swart 2010, Oliver & Smiley 2006, Yi 2005, 2006, Landman 2000, 1989a,b, Schwarzschild 1996, Link 1993, 1983, Krifka 1990, 1989, Schein 1986, Scha 1981, to name a few). The current discussion looks at a small part of this multi-faceted topic, namely the available readings of certain plural predicates. Gillon (1987) argues that certain plural predicates are ambiguous inrespect to the truth value of their minimal covers—i.e. sets of subsets ofpluralities, in which none of the subsets overlap with the sum of the others, and the sum of all subsets is equal to the plurality itself. While Gillonճ (1987) argument is logically sound, I present evidence that suggests that certain minimal covers are not immediately accessible interpretations of cumulative predicates. Namely, following certain plural predicates, the lexical modifiers together and individually cannot be used in tandem to indicate which minimal cover of the predication is true. Following Gillon (1987), I assume these minimal covers must exist despite their inaccessibility, so the issue at hand is determining how they come to exist and why they are inaccessible from the initial interpretation of certain plural predicates. Landmanճ (2000) analysis of events and plurality provides a multi-step process in which cumulative predicates are derived from covers. This analysis can be used to show how the interpretations of predicate that includes the relevant minimal covers come to exist, but leaves the inaccessibility of such covers unexplained. I propose that complex minimal covers are inaccessible because their derivation is simply too complex to process on-line. This explanation accounts for the infelicity of lexical modifiers, preserves the logic of minimal covers, and avoids the introduction of further devices into the account. |
| June 28, 2017 Mojmr Dočekal and Marcin Wągiel Masaryk University, Czech Republic Counting degrees and events: A cross-linguistic perspective In this talk, we bring in novel data concerning distribution and semantic properties of two classes of adverbs of quantification in Czech, i.e., event numerals such as *dvakrt* (‘twice/two times’) as opposed to degree numerals such as *dvojnsobně* (‘doubly/twofold’). We explore the contrasts between the expressions in question including the interaction with comparatives and equatives as well as scope asymmetries. Furthermore, we discuss their relationship with frequency adverbs and degree adverbs, respectively, as well as cross-linguistic variation including data from typologically distinct languages such as Vietnamese. We propose that degree numerals target values on a provided scale and are, hence, best analyzed as degree quantifiers resembling differentials whereas event numerals have a more general semantics which primarily allows for quantification over individuated events, but also enables to operate on degrees. |
| June 21, 2017 va Kardos University of Debrecen, Hungary Telicity across languages This talk is concerned with how telicity arises across languages. More specifically, it offers a semantic take on the encoding and calculation of telicity and identifies two types of strategies in typologically such different languages as English, Hungarian and Slavic languages. It promotes the idea that telicity arises either (i) as a result of overt or covert maximalization over events, as in the case of most predicates in Hungarian and Slavic languages, or (ii) simply due to the co-occurrence of a verb encoding incremental change, a theme participant whose quantity is known, and a bounded path that is traversed in the course of the denoted event, as with most English verbal predicates (Kardos 2012, 2016). Important ways in which the languages under investigation differ is that they rely on these strategies to different extents, which has significant interpretational and morphosyntactic differences with respect to verbal predicates. Following Filip and Rothstein (2006) and Filip (2008), I assume that telicity arises either as a result of a maximalization operator MAXE mapping sets of partially ordered events onto sets of maximal events. The application of MAXE is contingent on a verb assigning a figure-path incremental relation, a quantified theme DP, and a bounded path, an idea originating in Beavers (2012). Alternatively, telicity can also arise without maximalization over events with a verb assigning a figure path incremental role to a theme DP whose quantity is known and a bounded path since in these cases for any event that the verbal predicate describes, it does not describe any non-final subevent of that event. An important difference between the two processes is that the former leaves the predicate with quantized reference, and thus telicity is guaranteed, whereas in the latter case it is not a necessary consequence. This is illustrated in Hungarian and Slavic languages, where event maximalization is encoded in particle verbs and perfective verbs. In English, where event maximalization is not necessary for telic interpretations in the case of most predicates, degree achievements like warm and cool are well-known for aspectual variability (Hay et al. 1999). Furthermore, in the case of predicates containing a particle verb or a perfective verb, which arguably encode event maximalization, the internal argument is semantically constrained in a way that the quantity of its referent must be known, and this satisfies a necessary condition for telicity (see above). As for how the event maximalizing operator is encoded, there are different options to be explored. It is either the case that it is encoded covertly in VPs, as in some instances in English, or perfective verbs, as argued for by Filip (2008) in her analysis of Slavic languages, or that it is encoded overtly in perfective prefixes in Slavic languages or verbal particles in Hungarian. That particles and prefixes seem to systematically turn stative predicates into telic predicates in Hungarian and several Slavic languages may be taken as one piece of evidence for their being overt event maximalization/telicity markers. Yet another question that is worth pursuing is whether bounded events must be marked in Slavic languages similarly to Hungarian. If this requirement holds, the expectation is that predicates that are inherently bounded (e.g. achievements like die and break a vase and degree achievements associated with an endpoint like empty the fridge and straighten the rope) must contain some marking element (e.g. a particle or a prefix). In Hungarian, this expectation is met, whereas in Slavic languages different patterns arise, making obligatory telicity marking suspicious (cf. Di Sciullo & Slabakova 2005). |
| June 14, 2017 Łukasz Jędrzejowski University of Cologne On (the diachrony of) jakoby-clauses in Polish In this talk, I will examine the development and use of dependent clauses in Polish introduced by the complementizer jakoby (lit. ԡs ifթ and show which factors in the lexical meaning of jakoby were responsible for the semantic change that it underwent. In the Old Polish example given in (1), the dependent clause is introduced by the hypothetical comparative complementizer jakoby (ԡs ifթ and it is embedded under the matrix predicate widzieć (Գeemթ, expressing indirect inferential evidence: (1) ludziem na ziemi tako było widziało people.dat on earth.loc so be.l-ptcp.3sg.n seem.l-ptcp.3sg.n [1] jakoby się ono na nie obalić było chciało jakoby refl it on them slay.inf be.l-ptcp.3sg.n be.l-ptcp.3sg.n Դhe people on earth interpreted it as if it wanted to slay all of themռo:p> (KG, Kazanie I: Na Boże Narodzenie 26-7) In Old Polish, jakoby-clauses can be embedded only under verbs of seeming. In other words, the structure seem as if p is used instead of seem that p if what the available evidence suggests is somehow in conflict with what the speaker believes or used to believe. In Present-day Polish, in turn, as illustrated in (2), the jakoby‑clause is embedded under the speech verb zaprzeczać (Ԥenyթ: (2) Firma zaprzeczała, jakoby były company deny.l-ptcp.3sg.n jakoby be.l-ptcp.pl.n-vir [2] zgłoszenia o wadliwych kartach. reports about faulty cards.loc Ԕhe company denied that there supposedly were any reports about faulty prepaid cards.ՠ (NKJP, Dziennik Zachodni, 27/9/2006) The complementizer jakoby is not interpreted as a hypothetical comparative conjunction as if any longer, but as a hearsay complementizer (żi style=’mso-bidi-font-style:normal; mso-bidi-font-style:normal’>that + allegedly). Interestingly, neither Czech nor Slovak have experienced this change. Based on Faller (2011) and Murray (2017), I will present account showing that the change of jakoby involved two main developments: First, the meaning of jakoby was broadened to allow for inferences from reportative information (compatible with, but not enforced by its seem-type embedding verbs). Second, the reportative flavor acquired by jakoby licensed its use in complements of speech verbs. Since these new contexts were no longer compatible with the original inferential meaning, they ultimately lead to the inability to use jakoby in its original contexts, cf. (3): (3) *Firmie wydaje się, jakoby ɼo:p> company.dat seem.3sg refl jakoby Intended meaning: ԉt seems to the company as if ɕ References Faller, Martina (2011): A possible worlds semantics for Cuzco Quechua evidentials, in: Proceedings of SALT 20 ed. by Nan Li and David Lutz, eLanguage, 660‑683. Murray, Sarah E. (2017): The Semantics of Evidentials. Oxford: Oxford University Press. |
| June 7, 2017 Stergios Chatzikyriakidis University of Gothenburg In defence of lost causes? Type Theories for Natural Language Semantics In this talk, I will present an overview of the use of TTs for representing linguistic semantics. A historical overview is first given that covers in brief the history of type theory. Then, the discussion moves to the application of type theories, mostly type theories within the tradition of Martin Lf, to various issues in linguistic semantics like common nouns, modification, belief intensionality and copredication among others. Various alternatives are discussed when needed while the differences with simple type theory, which is the basis of Montague Grammar, are highlighted. Furthermore, the use of proof assistants implementing constructive type theories in dealing with Natural Language inference and checking the correctness of formal semantics accounts is discussed arguing that constructive type theories combined with the associated maturity in proof assistant technology can produce powerful reasoning engines and effective ҳemantic account checkersӮ Lastly, I discuss the use of TTs within the new setting in computational linguistics, namely Deep Learning, trying to see its usefulness or not, and potential avenues of interaction between the two fields. |
| May 31, 2017 Natalia Gagarina Leibniz-ZAS Berlin Aspect (and other verb categories) in monolingual and bilingual acquisition I trace the developmental projection of the acquisition of aspect (and other verb categories) in monolingual and bilingual children. Specifically, the goal is find out, how children use the verb morphology to gain access to syntactic structure and to identify differences in monolingual vs. (simultaneous and successive) bilingual acquisition. Working within the general framework of constructivism, I will show how the semantic structure of a predicate guides the acquisition of the tense/aspect and agreement morphology. Our methodology is called predicate tracking. Working within the lexicon of individual children, a predicate is identified and then the emergence of the verb morphology is tracked. Thus, the history of acquisition for each predicate is determined. Minimal morphological contrasts as evidence for the productivity of tense/aspect and agreement concepts are searched for. Prior research has shown a relatively consistent frequency pattern cross-linguistically where the following two configurations are highly probable: 1) telic verbs with past tense and bounded aspectual morphology, and 2) atelic verbs with non-past tense and unbounded aspectual morphology. This finding has prompted the argument that aspect emerges prior to tense in child language. According to some variations on the Principle and Parameters theme, this sequence is required by the principle of economy. Our research with the predicate tracking methodology indicates the following: 1) the precise pattern of acquisition is determined by the properties of lexical aspect, i.e., the logical structure of predicates, 2) the pattern varies cross-linguistically, and 3) deictic tense is likely to be productive prior to viewpoint aspect. |
| May 24, 2017 Leda Berio, Anja Latrouite, Robert Van Valin, Gottfried Vosgerau Heinrich Heine University Dsseldorf Immediate and General Common Ground The traditional literalist account of meaning has been challenged by several theories that stress the importance of context and of contextual information in communication, especially for mechanisms of meaning determination and reference fixing. However, the role of lexical meaning in such contextualist accounts often remains only vaguely defined. In this paper, we defend an account of communication that keeps the advantages of contextualist theories, while a new element is introduced that we claim could help solving some of the remaining issues. By differentiating Immediate and General Common Ground in communica- tion, we draw a distinction between mechanisms related to the situation at hand and those concerned with world and language knowledge. We further argue that such a distinction can help understanding cases of loose use and metaphors of which we provide some examples. Finally, we claim that this distinction has grammatical reality, as it is shown by the examples from Lakhota (North America), Umpithamu (Australia), Kuuk Thaayorre (Australia) and Mongsen Ao (India) discussed in the paper. |
| May 17, 2017 George Tsoulas University of York How to pluralise a Mass noun: The ingredients The relation between mass terms and plurals is well known and well supported empirically in terms, for example, of determiners that are common to the two classes of nominals to the exclusion of the singular. At the same time it comes in general as a surprise that given the notion that mass terms as “somehow plural” they never show up in the plural. From this point I look at a set of languages which do pluralise mass nouns and attempt to derive the possibility of plural mass terms. The objective is to build a syntactic and semantic framework with as minimal assumptions as possible, within which they can naturally occur. More specifically I look a the set of required morphosyntactic features, their position in the nominal extended projection and their associated semantics. Then we try to understand where the points of variation lie and what sort of variation is predicted. |
| May 3, 2017 Stefan Heim Uniklinik RWTH Aachen & Forschungszentrum Jülich If so many are “few”, how few are “many”? The neurocognition of quantifier processing The processing of quantifiers such as ҭanyӠor Ҧewӊ is a complex operation, involving the estimation of the numerosities of objects, their comparison to a reference amount, and semantic evaluation of that comparison. This series of processing steps is supported by a fronto-parietal network predominantly in the left hemisphere. The criterion that defines a number of objects as e.g. ҭanyӠdepends on the context (many pandas vs. many ants) and on personal experience (ҭany milesӠfor a long-distance runner vs. a person using crutches with a broken leg). I will report that this internal criterion can be modified in the course of a learning paradigm in which healthy young subjects can be trained to adapt their judgement of ҭanyӠfrom 60% to 40% of all circles. Most interestingly, changing the criterion for the quantifier ҭanyӠalso leads to a change in the criterion for the untrained quantifier ҦewӠ(which contracts to about 25%). Brocaճ region in the left inferior frontal cortex is essential for this learning effect and the generalization. This leads to the question of performance in patients with the behavioral variant of fronto-temporal dementia (FTD) who suffer from atrophy of their frontal cortices: Are they impaired in their semantic generalization capacity tested with this paradigm. To this end, FTD patients were compared to elderly healthy controls. The healthy controls learned the new criterion for “many”, and this also affected their criterion for “few” even though the criterion for “few” had not been trained. In contrast, the FTD patients also showed a learning effect for the new criterion trained for the quantifier ҭany,Ӡbut failed to generalize this criterion shift to the other quantifier ҦewӮ In line with the previous studies, these findings point at the central role of the left frontal cortex for semantic generalization. Since the patients were still able to perform the task and showed learning of ҭanyӠto direct feedback, the data suggest that the left frontal cortex is relevant for generalization of the meaning of semantic categories as an instance of cognitive flexibility. This generalization process, rather than initial learning, seems much more vulnerable to frontal degeneration. |
| April 26, 2017 Kathrin Byrdeck & Kurt Erbach Heinrich Heine University Dsseldorf Object Mass Nouns in Japanese Classifier languages are commonly assumed not to have a grammaticized mass/countdistinction among nouns. In this talk we pursue two main goals. First, based on data from Japanese and some insights in Sudo (2016, forthcoming), and Inagaki & Barner (2009), we provide new empirical evidence that Japanese nouns for concepts like BOOK, DOG, SHOE (encoded by prototypical count nouns in languages with a grammaticized mass/count distinctions) and collective artifacts like MAIL, FURNITURE have individuated counting bases. We provide empirical evidence in support of the claim that Japanese nouns like yūbinbutsu (`mailթ have the hallmark properties of object mass nouns: (i) they are compared according to cardinality in `more thanՠconstructions (also Inagaki & Barner 2009), and (ii) they are infelicitous in constructions associated with count nouns. To the extent that the existence of object mass nouns can be established in the grammar of Japanese, as we argue, it follows that Japanese has grammatical reflexes of the mass/count distinction, rather than just exhibiting cognitive and grammatical reflexes of individuation (atomicity) alone. Second, we offer a new formal analysis of the Japanese mass/count distinction, building Sutton & Filipճ (2016) context-sensitivity driven account of the mass/count distinction, and specifically of object mass nouns. Our results bolster a nascent growing body of studies strongly suggesting that Japanese (and other classifier languages) in fact do have direct grammatical reflexes of the mass/count distinction. |
| April 19, 2017 Emar Maier University of Groningen Eventive vs. Evidential Speech Reports In this talk I argue for a distinction between eventive and evidential speech reports. In eventive speech reports the at-issue contribution is the introduction of a speech event with certain properties. Typical examples include direct and free indirect speech. In evidential speech reports, by contrast, the fact that something was said is not at issue, but serves to provide evidence for the reported content. Typical examples include Quechua and Bulgarian reportative evidentials, Dutch and German reportative modals (‘schijnen’, ‘sollen’), and the German reportive subjunctive. Following up on an observation by Von Stechow & Zimmermann (2005:fn.16), I argue that English indirect discourse is ambiguous. In the current framework this means it allows both an eventive reading, where a reported speech act is at issue, and an evidential reading, where it is backgrounded. |
| January 11, 2017 Dolf Rami Georg-August Universitt Gttingen Names, Pronouns and Demonstratives as use-sensitive expressions Kaplan famously distinguished between pure indexicals and true demonstratives. Nevertheless, he thought that these two kinds of expressions are only two subvarieties of the same semantic kind, namely the kind of context-sensitive expressions. I will argue that Kaplan underestimated the significance of his distinction and that it is in fact a distinction in semantic kind. In my opinion, there are two different sorts of expressions whose reference depends in different ways on the occasion of their use and which should semantically berepresented in different ways. Firstly, there are context-sensitive expressions. The reference of these expressions depends on certain objective factors that constitute a context of use. These expressions have a Kaplanian character that determines their referent relative to a context of use. Examples of this kind are expressions like ҉Ӭ ҮowӠand ҴodayӮ They all share the following feature: It is impossible that two different uses of an expression of these kind have different referents relative to the same context of use. Formally they can be represented, as Kaplan did, by functions from contexts of use to intensions. Secondly, there are use-sensitive expressions. These expressions are neither sensitive to any objective factors of a context of use nor can their linguistic meaning be conceive of as a character. These expressions can have anaphoric or referential uses. Some of them have bound and pragmatic anaphoric uses. The reference of such an expression is determined relative to a referential or pragmatic anaphoric use by accompanying referential intentions. Uses of use-sensitive expressions are individuated by means of their accompanying referential intentions. Formally use-sensitive expressions are represented as indexed expressions that are semantically interpreted by means of an index-sensitive assignment function. Different indexes correspond to different referential intentions. Co-referential use-sensitive expressions are semantically interpreted by means of indexed-assignment-functions with the same output. In this sense, the referent of a use-sensitive expression is determined by subject factors that are independent of the parameters of a context of use. The reference of use-sensitive expressions can additionally be constrained in some cases by the use-conditional meaning of such an expression. Use-conditional linguistic meanings can either be captured (a) as restrictions on indexed-assignment-functions or (b) as restrictions on the class of adequate possible uses of an expression. It is possible that any two different uses of an expression of these kind have different referents relative to the same or different contexts of use. Some use-sensitive expressions also have bound uses and their corresponding assignment functions can be shifted by the right sort of quantifiers. Bare demonstratives are use-sensitives expressions without an additional use-conditional meaning and they only have pragmatic anaphoric uses. Proper names are use-sensitives expressions with an additional use-conditional meaning that restricts the possible referents of a name to its bearer; but names only have pragmatic anaphoric uses. Third person personal pronouns are use- sensitives expressions with an additional use-conditional meaning that restricts the possible referents of them to male or female individuals and they have both pragmatic and bound anaphoric uses. |
| December 14, 2016 Ruth Kempson King’s College London Language as Mechanisms for Interaction: the challenge of modelling dialogue My task is to introduce and justify the formalism of Dynamic Syntax (DS), whose central claim is that natural languages are mechanisms for interaction, defined by grammars that articulate the online incremental process of growth of interpretation/linearisation underpinning both production and parsing. The talk will then argue that with this perspective, there is evidence that language evolution could have been possible without having to presume rich innateness, sudden-switch change, or prior availability of higher-order inferential capacities such as mind-reading. My starting point is to illustrate data from conversational dialogue which are a huge challenge for syntax and semantics models, since sentence-based grammars are not at all well suited to express the facts of dialogue – and even for pragmatists, since dialogue dynamics provide evidence that successful communication does not have to involve reading othersՠminds or grasping some propositional understanding to be shared. Rather, what is essential is speaker/hearer interaction. Then I will give a sketch of Dynamic Syntax sufficient to show how the dynamics of dialogue will emerge as an automatic consequence of the framework as well as expressing universal constraints on the process of structural growth. Finally I shall show how the individual mechanisms that underpin anaphora, ellipsis, discontinuity effects, and scope dependencies, can all be seen to be vehicles for interaction, in virtue of generalisations across these phenomena which can only be expressed in dynamic, interactive terms. |
| December 7, 2016 Kurt Erbach Heinrich Heine University Dsseldorf Bare singular nouns in Hungarian and the mass/count distinction I argue for an analysis of Hungarian in which notional singular count nouns are semantically number neutral, and thus felicitous, with measure constructions and the WH-quantifier mennyi (Էhat quantity ofթ. This provides an alternative analysis to Schvarcz (2016) and Schvarcz & Rothstein (2015) who analyze the majority of Hungarian notional count nouns as dual life—i.e. mass or count depending on the context. They assume knyv (Ԣookթ is mass with mennyi & measure constructions, but it is count in cardinality constructions. However, certain Hungarian constructions indicate bare singular count nouns are interpreted as number neutral. Furthermore, measure constructions sanction the occurrence of mass and bare plural count nouns but disallow singular count nouns (Krifka 1989, Landman 2016). Following Landman (2016) allows a more straightforward analysis that shows the Hungarian nominal system is more like other mass/count languages has been previously thought. |
| November 30, 2016 Todor Koev Heinrich Heine University Dsseldorf Adverbs of Change, Aspect, and Anaphoricity Adverbs of change, such as quickly or slowly, have been known to give rise to a number of interpretations. A sentence like Kazuko ran to the store quickly can describe the intensity of the described action (a manner reading), the temporal extent of the entire event (a duration reading), or the time between the culmination of the event and some previous event (an anaphoric reading). It has also been noticed that available interpretations are sensitive to thelexical aspect of the verbal predicate; for example, The police quickly spotted the suspect is only compatible with an anaphoric reading for quickly. Existing accounts of adverbs of change (e.g. Cresswell 1978; Rawlins 2013) take manner readings as primary and successfully extend these to duration readings, but struggle to derive anaphoric readings. In contrast, I take anaphoric readings as primary and argue that other readings are special cases of these. Adverbs of change are claimed to modify the temporal distance between two instantaneous events that are compositionally or anaphorically available. The proposed account, couched in a dynamic semantic framework, does particularly well in predicting anaphoric interpretations and demonstrates how adverbs of change interact with lexical aspect to derive other available readings. It also correctly predicts certain positional effects and can accommodate the idiosyncratic behavior of adverbs like slowly, which appears to lack truly anaphoric interpretations. |
| November 23, 2016 Maria Spychalska Ruhr University of Bochum Scalar implicatures in context of full and partial information: Evidence from ERPs A major part of the psycholinguistic research on scalar implicatures has been focused on the question of how scalar implicatures are generated: in a default and automatic manner or as results of effortful reasoning processes. This so-called default- vs. context-based controversy has been experimentally operationalized in terms of processing costs of scalar implicatures: the processing costs have been taken as a proxy of the implicatureճ default vs. non-default character. Yet, it has eventually become evident that the data hardly fit this dichotomy. Many studies on the processing of scalar implicatures brought contrastive results: some experiments provided evidence that the processing of the pragmatically enriched interpretation is costly relative to the processing of the semantic meaning, other studies found no additional cost for the processing of scalar implicatures. It was further shown by Degen & Tannenahus (2015) that scalar implicatures may be differently processed depending on contextual support: in contexts that support the pragmatic interpretation, scalar implicatures will occur as default and automatic, whereas when the contextual support is weaker, listeners will take longer to arrive at the inference. This results were integrated within a probabilistic model of linguistic processing, called constraint-based account and predicting that interlocutors may use information from multiple sources during sentence comprehension to create expectations about the future development of the utterance. In my talk I will present results from EEG studies on scalar implicatures processing arguing in favor of the constraint-based model. Comparing results from two studies: where the scalar implicature processing was tested in context of full information and in context of partial information, I will discuss how the contextual support may determine the cognitive costs of the implicature processing. |
| November 16, 2016 Willy Geuder Heinrich Heine University Manner adverbs, agentive adverbs, and adverbs in between I would like to discuss with you some questions that relate to the distinction of different semantic adverb types and their frame-based analyses. First, I present a brief sketch of a frame-semantic approach to manner modification, which allows us to distinguish manner adverbs from other types of event-related adverbs, hence going beyond the characterisation of manner adverbs as “predicates of events”. Then, I discuss so-called “agentive adverbs” (Geuder 2002) (like “stupidly” in “The defender stupidly passed back”), arguing that they may be seen essentially as “predicates of events”, too, but involving a different relation to an e-variable, compared with manner adverbs. I explore the possibility that their non-restrictive function, agent orientation and scope-taking behaviour can be understood on the basis of their lexical meaning as “abstract” properties of events, i.e. involving a constitution relation between concrete and abstract event descriptions, inspired by recent work of Sb (2016). |
| November 9, 2016 Christian Wurm Heinrich Heine University The algebra of ambiguity We present an algebraic approximation to the semantic content of linguistic ambiguity. Starting from the class of ordinary Boolean algebras, we add to it an ambiguity operator and a small set of (rather peculiar) axioms which we think are correct for linguistic ambiguity beyond doubt. We then show some important, non-trivial results that follow from this axiomatization, which turn out to be surprising and not fully satisfying from a linguistic point of view. The results leave us with some open questions, both on the linguistic algebraic side – like the nature of intention in ambiguous statements, or the properties of disjunctive axioms in universal algebra. |
| July 20, 2016 Barbara Tomaszewicz Universitt zu Kln Focus association in sentence processing The focus structure of a sentence reflects the discourse context, but in the presence of various operators, such as only, even and most, it has an effect on the truth-conditions or the presuppositions of the sentence. Stolterfoht et al. (2007) and Carlson (2013) showed that only facilitates the processing of focus structures during silent reading. When (1) is read without preceding context, the first conjunct receives a wide focus interpretation (marked as FI), and when the processor encounters the ellipsis remnant (F2), it must revise the focus structure of the first conjunct from wide to matching narrow focus (F3). The presence of only in (2) requires narrow focus on its associate (FI), which is congruous with the ellipsis remnant. Revision of the focus structure in (1) vs. (2) was associated with an ERP signature in the Stolterfoht et al. study on German, and with increased reading times in Carlsonճ self-paced reading study on English. 1) [Am Dienstag hat der Direktor [den Schler]F3 getadelt]F1, und nicht [den Lehrer]F2 On Tuesday has the principal.Nom the pupil.Acc criticized and not the teacher.Acc 2) Am Dienstag hat der Direktor nur [den Schler]F1 getadelt, und nicht [den Lehrer]F2 On Tuesday has the principal.Nom only the pupil.Acc criticized and not the teacher.Acc Since ԯnly x … and not yՠis frequent in discourse, the presence of only could create an expectation for an explicit mention of excluded alternatives, and this bias alone could account for the facilitation in (2). In self-paced reading experiments on Polish we showed that the processing of replacive ellipsis ( ԡnd not … ՠ) is facilitated in the presence of the three associators: only, even, most, which indicates that it is indeed the focus association mechanism that explains the facilitation in (2) (Tomaszewicz and Pancheva (2016)). The use of Polish allowed us for a direct comparison between only, even and most, because (i) like in German replacive ellipsis is unambiguous due to Accusative case marking, and therefore any differences in ellipsis resolution can be attributed to the processing of focus structure alone; (ii) with most the focus on Գculptorsՠyields a superlative reading that is unavailable in English or German (in Pancheva and Tomaszewicz (2012) and Tomaszewicz (2015) we argue that this reading arises via focus association). We found that in Polish only, even and most create an expectation for narrow focus on the object, but the facilitatory effects occur already on the conjunct ԡnd notՠwith only and even, and on the ellipsis remnant with most. This difference likely reflects the difference between two types of focus association: obligatory and optional as identified in the formal semantic research on focus. Obligatory focus association is taken to be encoded in the lexical semantics of focus sensitive expressions (only, even), whereas optional/free association is a result of the contextual setting of the domain variable of an operator like most (Beaver and Clark (2009)). While prenominai only and even have one syntactic associate (3), most is free to associate either with the adverbial or the subject in English (4a-b), or with the object in Polish (5). 3)a. John invited only/even [sculptors]F for coffee. b. *John invited only/even sculptors [for coffee]F 4)a. John invited the most sculptors [for coffee]F. Reading: John invited more sculptors for coffee than for any other relevant occasion, b. [John]F invited the most sculptors for coffee. Reading: John invited more sculptors for coffee than for any other relevant individual did. 5) John zaprosił najwięcej [rzeźbiarzy]F na kawę. John invited most sculptors for coffee Reading: John invited more sculptors for coffee than any other group of people that he invited. During incremental processing prenominai only and even create a precise expectation for the location of focus, but most allows association with either the object or the adverbial in Polish, which is compatible with our results. Currently, we are extending these findings to meisten in German, which like English most does not allow association with the object, to show that optional associators facilitate the processing of focus structures that are compatible with the semantics resulting from focus association (and that it is not the case that the mere presence of a prenominai modifier increases the salience of the contrast in the replacive ellipsis). |
| July 6, 2016 Noortje Venhuizen Universitt des Saarlandes Projection in Discourse: A data-driven formal semantic analysis In this talk, I present a unified, data-driven formal semantic analysis of projection phenomena, which include presuppositions, anaphoric expressions, and conventional implicatures (as defined by Potts, 2005). The different contributions made by these phenomena are explained in terms of the notion of information status. Based on this analysis, I present a new semantic formalism called Projective Discourse Representation Theory (PDRT). PDRT is an extension of traditional Discourse Representation Theory (Kamp, 1981; Kamp and Reyle, 1993), which directly implements the anaphoric theory of presuppositions (van der Sandt, 1992) by means of the introduction of projection variables. I show that PDRT captures the differences, as well as the similarities between the contributions made by presuppositions, anaphora and conventional implicatures. In order to illustrate PDRT’s representational power, I present a data-driven computational analysis of the information status of referential expressions based on data from the Groningen Meaning Bank; a corpus of semantically annotated texts (Basile et al., 2012). Taken together, the results pave way for a more integrated formal and empirical analysis of different aspects of linguistic meaning. |
| June 29, 2016 Markus Schrenk Heinrich-Heine-Universitt Dsseldorf Causal Power and Modality as Stumbling Stones for a Semantic Analysis of Dispositional Predicates Attempts to give a semantic analysis of dispositional predicates (like ҳolubility”, ҩnflammabilityӬ etc.) in terms of (counterfactual) conditionals (for example: ҩf x were put in water it would dissolveө saw a plethora of counterexamples: void satisfaction, random coincidences, masks, finks, antidotes, etc. Already on the (semantic/logical) surface the reasons for failure are pretty obvious. Yet, there are also some deeper (metaphysical) reasons – Causal Power and Modality – that can be unearthed. In this talk I will go through all the above. |
| June 22, 2016 Ekaterina Rakhilina Higher School of Economics, Department of Linguistics, Moscow A Typology of Falling Events Falling is a kind of quite standard motion event with Source, Goal and a special manner of motion which is usually reduced to the up-down (vertical) axis. The talk shows that there are other semantic characteristics which build semantic oppositions within the domain of falling and trigger its metaphorical extensions. |
| June 15, 2016 Laura Kallmeyer & Behrang Qasemi Zadeh Heinrich-Heine-Universitt Dsseldorf The PoP approach to vector semantics Vector semantics represents the meaning of a word by characterizing its distribution. More concretely, word meanings are represented by distributional vectors where the dimensions of these vectors denote context words and the coordinates are determined by the frequency of these context words in the neighbourhood of the word that the vector characterizes. Notions of similarity and distance between vectors can then be used to infer similarity of meaning between words. The use of vector space models, however, is problematic due to the high dimensionality of these vectors and due to the fact that co-occurrence patterns often follow what is known as heavy-tailed distribution (as exemplified by the Zipfian distribution of words in documents). For instance, while a few words frequently occur in text (thus co-occur with other words, such as the function word Ҵheө, many content words occur rarely. Consequently, as the number of vectors/entities increases, the number of co-occurring context elements (i.e., the dimensionality ofvectors) escalates. In this talk, we first describe principles of vector semantics including the above-mentioned problems. We then introduce a new technique called positive-only projections (PoP) that address the problem of high dimensionality. PoP allows to build vectors at a fixed reduced dimensionality and in an incremental fashion. We report the performance of PoP method in two semantic similarity measurement tasks: TOEFL synonym test and MEN relatedness. In both tasks, PoP shows a performance comparable to state-of-the-art neural embedding techniques. |
| June 8, 2016 Suzi Lima Universidade Federal do Rio de Janeiro Portions, individuation and measurement Container nouns (cup) are nouns that denote concrete objects that can be used as receptacles for substances. It has been argued that in constructions with numerals (as in Դwo glasses of waterթ, container phrases can be interpreted in at least two different ways (Selkirk 1977, Rothstein 2012, Partee and Borschev 2012). Firstly, a container noun can be used to denote actual containers filled with some substance; e.g. ԧlasses of waterՠcan denote actual glasses filled with some quantity of water (individuation). Secondly, a container noun can be used as the description of a unit of measurement. Based on the results of a felicity judgment task with children and adults with Brazilian Portuguese, English and Yudja, it will be argued that individuation precedes measure. Second, unlike in English, container phrases in Yudja can be interpreted as locatives or trigger a concrete portion interpretation that is different from measure. Similar results were found for Brazilian Portuguese when the question included a prepositional phrase (Eu bebi dois copos com gua ԉ drank two cups with waterթ as opposed to pseudopartitive constructions (Eu bebi dois copos de gua ԉ drank two cups of waterթ. |
| June 1, 2016 Jens Fleischhauer Heinrich-Heine-Universitt Dsseldorf Animacy and affectedness (in Germanic languages) Most Germanic languages (English is an exception in this regard) show an animacy-dependent marking alternation of the second argument of contact verbs such as hit, kick or pinch. This is illustrated by the German examples in (1). If the referent of the second argument is inanimate, it is marked by a preposition (1b). (1) a. Das Mdchen schlug den Jungen. the girl hit the boy Ԕhe girl hit the boy.զnbsp; b. Das Mdchen schlug *(gegen/auf) den Tisch. the girl hit against/on the table Ԕhe girl hit against/on the table.զnbsp; The alternation shown in (1) is not solely dependent on animacy but also on affectedness. If the referent of the inanimate is definitely affected by the contact – for example in case of a resultative construction – , it is not marked by a preposition (2). (2) Das Mdchen schlug (*gegen/auf) den Tisch in Stcke. the girl hit against/on the table in pieces Ԕhe girl hit the table in pieces.զnbsp; Lundquist & Ramchand (2012) argue that inanimate entities are conceived as less affected by processes such as hitting or kicking than animate entities are. de Swart (2014), on the other hand, argues that the alternation marks a difference in sentience. As sentience presupposes animacy, the animacy contrast is merely epiphenomenal. Both analyses have shortcomings: de Swartճ analysis does not rely on affectedness and therefore cannot explain the contrast between (1b) and (2). Lundquist & Ramchandճ analysis is couched in the generative framework and they define affectedness as a binary feature. Their analysis does not give a principal explanation of why it is only a subset of contact verbs that gives rise to the alternation illustrated in (1). An explanation of them phenomenon requires two things: First, a graded concept of affectedness, like the one proposed by Beavers (2011). Beavers notion of affectedness provides an explanation of why contact verbs show an alternation dependent on affectedness. According to him, these verbs entail potential results and the alternation can be seen as a resolution of this potentiality. If the referent of the second argument is animate, it is conceived as affected. If it is inanimate, it is taken to be non-affected. Second, an explication of the relationship between affectedness and animacy is needed. Lundquist & Ramchand argue that inanimate entities are only affected, if they are physically damaged. Beside physical affectedness, animate beings can also be emotionally/psychologically affected. This allows combining the basis insights of Lundquist & Ramchandճ analysis with the one of de Swartճ. The aim of the talk is to present a unified analysis of the phenomenon, which combines a gradual notion of affectedness with the notion of sentience. It will be shown that such an approach allows explaining why the alternation arises with this particular set of verbs. Furthermore, the analysis will shed light on the relationship between affectedness and animacy. |
| May 25, 2016 Karoly Varasdi & Zsofia Gyarmathy Heinrich-Heine-Universitt Dsseldorf A model of evidence and an evidence-based analysis of progressive achievements In our talk, we are going to outline a lattice-theoretic model of evidence and an evidence-based approach to progressive achievements. Our starting point is that an agent is justified in asserting a progressive sentence if the agent has enough evidence supporting the base sentence in that specific scenario. In our framework, evidence for a proposition is that which justifies the speaker in asserting the proposition. We will argue that the set of all potential pieces of evidence ordered by containment form a lattice, and is connected with the lattice of propositions in a specific way based on the notion of (partial) justification. We will show how this evidence-based framework can be used to predict felicitous progressive uses of achievements, such as “Mary is arriving at the station” and still exclude unacceptable progressive achievements like “*Mary is noticing the picture”. To this end, we exploit the fact that sentence entailment and evidence containment go in opposite directions in our framework and that achievements that can appear in the progressive are those that describe the right boundaries of extended events. |
| May 4, 2016 Zsofia Gyarmathy Heinrich-Heine-Universitt Dsseldorf Achievements and presuppositions This session will include i) a presentation of a theoretical idea about the activity presupposition of a subclass of Vendler’s achievements (called culminations by Bach 1986), ii) outlining some ideas for three experiments to test the theoretical assumptions, and iii) an open discussion where I would be happy to hear comments and feedback especially about the experimental design. The basic theoretical idea I propose is that culminations like win or arrive have an existential presupposition that is of a hitherto not recognized kind that I call an extra soft presupposition, which is cancellable even when the trigger is not embedded under any operators, but is more similar to presuppositions than implicatures in other respects. |
| April 27, 2016 Daniel Lassiter Stanford University Time: 16:30-18:00 Location: Building 24.91. Room 01.22 Must, might, knows, and the rest of the epistemic system Linguists and philosophers have long been torn between the intuition that must is ‘weak’ – expressing reduced commitment vis-a-vis the unqualified expression – and the intuition that it expresses some fairly strong epistemic relation, such as knowledge. Recently von Fintel & Gillies (2010) have argued that the latter hunch is correct – must picks out a strong epistemic necessity modal la modal logic S5 – and that the intuition of weakness can be explained by reference to a little-noticed evidential meaning component of must. Using corpus and experimental data I’ll show that must does not express knowledge, certainty, or anything of this form: speakers routinely use must to mark out a proposition p when they are explicitly uncertain about the truth of p, say that they do not know p, and consider not-p a possibility. The experimental results also illuminate the relationship between might and epistemic possible, which are (contrary to the usual assumption) not synonymous. I’ll discuss the implications of these results for a variety of epistemic items, arguing that they problematize Kratzer’s (1991) influential proposal as well and favor a theory where epistemic modals are given a semantics built around the probabilistic support of a proposition. |
| April 21, 2016 Todor Koev Heinrich-Heine-Universitt Dsseldorf Countability and the Diminutive in Bulgarian This technical report briefly explores the count/mass distinction in Bulgarian. It pays particular attention to diminutive modification and its ability to achieve a mass-to-count shift when applied nouns that describe granular aggregates. |
| April 14, 2016 Michael Daniel National Research University, Moscow Higher School of Economics Centre for Fundamental Studies / Laboratory of the Caucasian Languages Mass and Class: Number, nominal classes and mass nouns in East Caucasian In many ways, East Caucasian languages manifest a complex interaction between the categories of grammatical number and gender as nominal classification. One obvious example of such interaction is that the inventory of nominal classes in the singular (usually three to five classes) is reduced to only two classes in the plural – human vs. non-human. Not less important, though probably somewhat less salient, is the treatment of mass nouns as P(luralia) T(antum). Thus, in Dargwa languages, mass nouns show non-human plural agreement. That mass nouns are PT is, of course, by no means typologically unexpected. What is peculiar is that the mass nouns show this agreement though being morphologically singular, while, at least for some of them, morphologically plural forms are also available. Similar but more complex is the situation in Archi, an outlier of the Lezgic branch of the family. In Archi, mass nouns are usually described as fourth class (singular). Incidentally, agreement pattern for this class is identical to non-human plural. Many of these nouns show plural inflectional morphology. This morphology is however not consistent and is coupled with singular agreement on the attributes. I suggest that mismatches between morphology and agreement are explained by a more general morphosyntactic property of East Caucasian languages that do not ascribe agreement pattern to a lexical item as a whole but do this separately to its singular and plural forms, and that the ascription is governed not only by lexicon but is partly driven by semantics of the respective morphologically singular and plural forms. |
| February 10, 2016 Simon Dobnik University of Gothenburg Interfacing Language, Spatial Perception and Cognition in Type Theory with Records In the proposed presentation we overview and connect two lines of our work related to Type Theory with Records (TTR): modelling of spatial language and cognition and modelling of attention-driven judgement. We argue that computational modelling of perception, action, language, and cognition introduces several requirements on a formal semantic theory and its practical implementations: (i) interfacing discrete conceptual knowledge and continuous real-valued sensory readings; (ii) information fusion of knowledge from several modalities; (iii) dynamic adaptation of semantic representations/knowledge as agents experience new situations through linguistic interaction and perception. Using examples of semantic representations of spatial descriptions we show how Type Theory with Records satisfies these requirements. The advantage of truth being based on agent-relative judgements in TTR is crucial in this but practically it comes with a computational cost. However, this challenge is not unique to TTR. An agent would have to check whether a situation s is of every type in its inventory. In the second part of the talk we argue that the number of type judgements an agent has to make can be minimised by incorporating a cognitive notion of judgement that is driven by perceptual attention. |
| January 27, 2016 Fabienne Martin Stuttgart University The imperfective in subjunctive conditionals: fake or real aspect? This talk aims to provide a ‘real aspect’ approach of the ‘fake’ imperfective in subjunctive conditionals (SCs) and a new account of the (non)-cancellability of the counterfactual inference in SCs, largely based on Ippolitoճ 2013. It is argued that PAST and PRES above MODAL in conditionals compete the same way as in non-modal stative sentences, see Altshuler and Schwarzschild 2012. On this view, the counterfactual inference of SCs, when cancellable, is nothing else than the cessation implicature routinely triggered by past stative sentences. |
| January 20, 2016 Fred Landman Tel Aviv University and University of Tbingen Aspects of Event Semantics for Aspect Note that this is more a tutorial on aspects of my work on event semantics than a lecture. Topics: 1. Event models based on a discourse pragmatic notion of cross-temporal identity of events and the notion of event stages. 2. Homogenenous eventuality types. Homogeneity for statives, Incremental homogeneity for activities and the semantics of ‘for an hour’. 3. Telicity, stabalization, and the semantics of ‘in an hour’. 4. Stativity operators: a new proposal for the perspective operators of “1066” paper). 5. Scalarity and the perfect/progressive (in “1066” paper) |
| January 13, 2016 Todor Koev Heinrich-Heine University Parentheticality and Discourse Sentences with slifting parentheticals (e.g. *The dean, Susan said, flirted with the secretary*) span the divide between semantics and pragmatics because in them the main clause plays a central role while the slifting parenthetical describes the grounds for asserting the main clause (cf. Urmson 1952; Hooper 1975; Asher 2000; Rooryck 2001; Jayez & Rossari 2004; Davis et al. 2007; Simons 2007; Scheffler 2009; Haddican et al. 2014). In this talk, I discuss three core properties of sentences with slifting parentheticals: (i) the backgrounded status and projection/scopal properties of implications triggered by slifting parentheticals, (ii) the often weakened assertion strength of the main clause, (iii) the requirement that slifting parentheticals create an upward-entailing environment (cf. #*The dean, Susan doubts, flirted with the secretary*). I will argue against the idea that the main clause is interpreted in the scope of the slifting predicate. Rather, I suggest that the strength of the main clause depends in a quasi-pragmatic way on the slifting parenthetical, as the latter can lower the assertability threshold for the main clause. I will also try to derive the informational properties and the polarity restrictions on slifting parentheticals from their role as providing grounds for the main claim of the sentence. |
| December 9, 2015 Paul Gaus Heinrich-Heine University Result States in the Perfect Time Span – Combination of two Theories of the Perfect The Perfect Time Span approach (PTS) is an approach which tries to capture perfect meanings by locating the Reichenbachian Event Time (E) within a time span. I found that in German certain accomplishment constructions cannot be captured by this approach. In this thesis I will show a solution to this problem by combining the PTS approach with the so called Result State approach. My modi cation of the PTS approach no longer predicts E within a time span but the onset of the result state of the eventuality which takes place at E. This allows the problematic construction to be predicted, in addition to the non-problematic constructions. Further I will discuss the result state approach and show that this approach also needs time spans to predict certain meanings which brings me to the claim that a combination of both theories is desirable because it has a better empirical coverage. |
| December 3, 2015 Todor Koev Heinrich Heine University Appositive Projection and Its Exceptions This paper has two major goals. The first is to offer a comprehensive account of the projection properties of appositive constructions. Appositives posit a challenge to traditional assumptions about form and meaning because they are interpreted in situ with respect to order-dependent phenomena like discourse anaphora but nevertheless escape the scope of entailment-canceling operators like negation or modals. Accounting for this pattern requires an innovative way of looking at propositional operators and how they interact with appositives. The second goal of the paper is to address various claimed exceptions to the otherwise robust projectivity of appositives. I argue that in some cases the construction under consideration is most likely not an appositive at all. In other cases, the observed non-speaker-oriented readings can be derived by pragmatic reasoning or are due to a perspective shift. Although genuine instances of semantically embedded appositives do seem to exist, I point out that such data have a limited empirical scope. I conclude that appositive projection is a pervasive phenomenon and is part and parcel of the semantics of appositives. |
| November 25, 2015 Peter Sutton and Hana Filip Heinrich Heine University Dsseldorf Mass/Count Variation: A Mereological 2d Supervaluationist Semantics We propose a novel analysis of the mass/count distinction, within a new framework: 2-dimensional mereological supervaluationism. While the notions akin to VAGUENESS [1], SEMANtiC ATOMICITY [4] and OVERLAP [3] are needed to ground this distinction, no single notion is sufficient to fit the whole range of data, especially intra- and crosslinguistic variation in mass [-C] vs. count [+C] encoding. We make this variation tractable by treating it as following from the interaction of all three of the above notions. We formally derive four semantic classes of nouns (which closely match those in [2]) to explain these form-denotation mappings and overcome challenges faced by [1; 3; 4]. References [1] Chierchia, G., 2010. Mass nouns, vagueness & semantic variation. Synth. 174, 99-149. [2] Grimm, S., 2012. Number & Individuation. PhD Diss., Stanford University. [3] Landman, F., 2011. Count nouns, Mass nouns, Neat nouns, Mess nouns. In: The Baltic International Yearbook of Cognition: Vol. 6. pp. 1-67. [4] Rothstein, S., 2010. Counting & the mass/count distinction. JoS 27 (3), 343-397. |
| November 19, 2015 Susan Rothstein Bar Ilan University (Israel) & Tbingen University Object mass nouns from a crosslinguistic perspective Rothstein 2010, Schwarzschild 2011 show that nouns like furniture denote sets of individuable entities. Barner and Snedeker (2005) show further that comparisons such as who has more furniture? typically are answered by comparing cardinalities. On this basis they suggest that object mass nouns have essentially the same denotations as count nouns. In the first part of the talk, I will show that the conclusions drawn by Barner and Snedeker (2005) are too strong: while comparisons of object mass nouns may involve comparing cardinalities, they need not do so. There is thus a basic contrast between object mass nouns and count nouns: count nouns require comparison by cardinality while object mass nouns allow it, but also allow comparisons along other, continuous dimensions. I will support this with data from English, Brazilian Portuguese, Hungarian and Mandarin. This means that object mass nouns and count nouns must have different semantic interpretations, contra e.g. Bale and Barner (2009). But whatever semantics we give for object mass and count nouns, we need to answer the obvious question: If object mass nouns are not countable, how can they be compared in terms of cardinality? In the second part of the talk, I offer a solution to this problem, proposing that there are cardinality scales, which allow us to evaluate and compare quantities in terms of their perceived or estimated number of atomic parts without actually counting the atoms. This allows us to clarify the distinction between counting and measuring, and to maintain the general principle that only count nouns have countable denotations. |
| November 11, 2015 Henk Zeevat Heinrich Heine University Presupposition Blocking by Causal Inference The talk develops an account of causal inferences in update semantics. A typical case would be the inference of a causal relation in: When John pushed the button, the bomb exploded. While cause inferences have many applications, the talk applies it to presupposition projection. It shows that the effects of Karttunen’s satisfaction theory are better captured by causal inferences than by the logical satisfaction relation employed by Karttunen. E.g. blocking is not predicted by satisfaction in: If Marie is French, she has stopped eating snails. It is after all just false that every French person eats snails. But there is a plausible causal inference that being French can lead to eating snails. In: If John has grandchildren, his children are happy. most people infer that John has children, contra the satisfaction theory. And of course, children are the cause of grandchildren and not inversely.The talk also adds a new class of presupposition blocking that is unreducible to the satisfaction theory, based on identity inferences. |
| November 4, 2015 Rainer Osswald Heinrich Heine University Quantification in Frame Semantics with Hybrid Logic |