Workshop: Approaches to the semantics of clause-embedding predicates: theories, cross-linguistic data, and experimentation
21 – 23 October / Online
Registration: Please fill out this form by 18 October to receive the zoom link for the workshop.
Programme: Thursday 21 October (all times in BST)
9:00-9:40 am | Sunwoo Jeong Melodies of Truth: Deriving Factive Inferences from Alternatives |
9:40 – 10:20 am | Kajsa Djärv & Maribel Romero Emotive factives and strong exhausitvity: an experimental investigation |
10:20 – 11:00 am | Paloma Jeretič Neg-raising modals and scaleless implicatures |
break | |
12:30 – 1:10 pm | Kajsa Djärv Embedded main clause phenomena: new perspectives on embedded illocutionary acts |
1:10 – 1:50 pm | Deniz Özyıldız Question embedding, neg-raising, and lexical aspect |
break | |
2:05 – 2:45 pm | Ciyang Qing From experimental semantics to scalable semantic fieldwork |
2:45 – 3:25 pm | Aaron Steven White A broad-coverage semantic classification of the English clause-embedding lexicon |
break | |
3:40 – 4:20 pm | Judith Tonhauser The Gradient Projection Principle: Further evidence and challenges |
4:20 – 5:00 pm | Shane Steinert-Threlkeld & Jakub Szymanik Explaining two semantic universals of clause-embedding predicates |
Programme: Friday 22 October (all times in BST)
10:00 – 10:40 am | Kristina Liefke Meaning-driven combinatorial restrictions and ‘imagine *whether’ |
10:40 – 11:20 am | Nadine Theiler Pragmatics intrudes into presupposition projection from wh-complements |
11:20 – 12:00 am | Jenny Doetjes Interpreting French wh-phrases in situ |
break | |
1:30 – 2:10 pm | Tom Roberts Parenthetically-embedded questions |
2:10 – 2:50 pm | Open slot: casual discussion |
break | |
3:05 – 3:45 pm | Donka Farkas & Floris Roelofsen Encapsulated Questions |
3:45 – 4:25 pm | Tanya Bondarenko When clauses are Weak NPIs: polarity subjunctives in Russian |
break | |
4:40 – 5:20 pm | Craige Roberts Meaning postulates in lexical pragmatics: Background vs. presupposition |
Programme: Saturday 23 October (all times in BST)
2:00 – 3:00 pm | Discussion: theories |
break | |
3:30 – 4:30 pm | Discussion: Experimental and computational methodologies |
break | |
5:00 – 6:00 pm | Discussion: Cross-linguistic variation and uniformity |
ABSTRACTS
Sunwoo Jeong
Melodies of Truth: Deriving Factive Inferences from Alternatives
Information structure, and one manifestation thereof, prosody, has been known to be some of the key factors that shape the projection behaviors of factive inferences. In this talk, I introduce less discussed patterns of prosody-sensitive factivity variation, and explore their implications to theories of factivity and alternative semantics. In the first half of the talk, I report on experimental findings which capture that in Korean, prosody and information structure systematically regulate the derivation of factive inferences not only in projective environments, but also in unembedded environments, i.e., in the absence of any entailment canceling operators. Furthermore, I show that the observed effect of prosody applies non-uniformly, and interacts in significant ways with verb type. To capture these patterns, I propose an analysis involving a new interpretive principle, which dictates that a given utterance be associated with a unique at-issue contrast between salient alternatives. In the second half of the talk, I explore the specific predictions of this account in more depth. In particular, the account is shown to generate distinct predictions for cases involving anaphoric contrasts, depending on which factors, among prosody, context, and the lexicon, play a primary role in determining the relevant alternatives. These additional predictions are tested empirically. The emerging patterns illuminate a complex interplay between focus, context, and lexical contrast in shaping prosody-sensitive factive inferences.
Kajsa Djärv & Maribel Romero
Emotive factives and strong exhausitvity: an experimental investigation
This talk investigates the question of whether so-called emotive factives (like resent, be surprised) differ from other types of factives (like know, forget) in terms of allowing for strong exhausitvity (SE) in sentences like It surprised Lisa who won. To-date, there are mixed empirical claims in the literature, with a number of researchers arguing that emotive factives only allow for weakly exhaustive (WE) readings (Berman 1991, et seq). However, recent experimental work by Cremers & Chemla (2017) finds evidence for SE readings with surprise (compared toknow and forget). We present results from an experimental study which extends this work in two ways: first, by testing a wider range of emotive and non-emotive factive predicates, and secondly, by manipulating the context as a way of investigating the source of SE readings. We find that all verbs tested show a small but significant ‘penalty’ in contexts where only the SE reading is true. However, there is no evidence that the emotive factives as a class differ from the non-emotives in this respect; thus suggesting that emotive and non-emotive factives do not differ from one another in terms of the availability of SE readings in the contexts we investigated. Interestingly, we also find strong evidence that know does not allow for WE readings.
Paloma Jeretič
Neg-raising modals and scaleless implicatures
Neg-raising behavior is a pervasive phenomenon among modals cross-linguistically: there are many apparent necessity modals which are interpreted as taking wide scope above negation, despite being expected to take narrow scope. I argue that the neg-raising behavior of certain modals can be captured by their ability to trigger ‘scaleless implicatures’. A scaleless implicature (SLI) is a polarity-sensitive and QUD-sensitive strengthening from an existential meaning to a universal one. The ability of an item to trigger a SLI is lexically encoded, but its particular behavior is sensitive to the presence or absence of scalemates in the lexicon of the language. SLIs can be straightforwardly analyzed in existing grammatical theories of scalar implicature (e.g. Fox 2007). I present a typology of modal SLI triggers, where they vary along two dimensions: underlying quantificational force, and whether they are ‘obligatory’ or ‘optional’. First, they can have either existential or universal quantificational force. Existential SLI triggers strengthen to a universal interpretation in unembedded contexts. When negated, they stay existential. There are also universal SLI triggers that, when negated, have the expression strengthened to the equivalent of a wide scope universal. In both cases, the item has a basic neg-raising profile. Second, there are ‘obligatory’ and ‘optional’ SLI triggers. In unembedded contexts and the right QUD conditions, obligatory SLI triggers must trigger a SLI, while optional SLI triggers can trigger either a SLI or a scalar implicature. Based on empirical evidence from various SLI triggers, I claim that this behavior depends on whether an appropriate scalemate is present in the lexicon of the language, under the right notion of what is an ‘appropriate scalemate’. I discuss how this finding affects our theories of scalar alternatives.
Kajsa Djärv
Embedded main clause phenomena: new perspectives on embedded illocutionary acts
In this talk, I examine the claim that embedded verb second (EV2), a well-studied ‘main clause phenomenon’, is associated with the illocutionary force of assertion. I argue that the illocutionary force of EV2-clauses is best captured in terms of the statement that they share the conventional discourse effects typically associated with main clause declaratives; in the sense of Farkas and Roelofsen (2017). I show that this proposal nicely captures the core discourse effects associated with EV2, while still allowing for a wide range of attested uses (based on naturally occuring data). I further show that the range of uses associated with EV2-clauses are not readily captured by accounts that link the availability of EV2 to an assertive operator or feature (e.g. Krifka 2014, a.o). The current proposal also has consequences for theories of clausal complementation and theoretical modelling in the framework of inquisitive semantics.
Deniz Özyıldız
Question embedding, neg-raising, and lexical aspect
Abstract
Ciyang Qing
From experimental semantics to scalable semantic fieldwork
Abstract
Aaron Steven White
A broad-coverage semantic classification of the English clause-embedding lexicon
I investigate which patterns of lexically triggered doxastic, bouletic, neg(ation)-raising, and veridicality inferences are (un)attested across clause-embedding verbs in English. To carry out this investigation, I use a multiview mixed effects mixture model to discover the inference patterns captured in three lexicon-scale inference judgment datasets: MegaVeridicality, MegaNegRaising, and MegaIntensionality, which capture veridicality, neg-raising, doxastic, and bouletic inferences across a wide swath of the English clause-embedding lexicon. I focus in particular on inference patterns that are correlated with morphosyntactic distribution, as determined by how well those patterns predict the acceptability judgments in the MegaAcceptability dataset. I find that there are 15 such patterns attested under this criterion—about 1% of those that are logically possible. I discuss how these patterns map onto morphosyntactic distribution and what their structure might imply about constraints on lexicalization.
Judith Tonhauser
The Gradient Projection Principle: Further evidence and challenges
The Gradient Projection Principle (GPP; Tonhauser, Beaver & Degen 2018, Journal of Semantics) maintains that a content C projects from an entailment-canceling environment to the extent that it is not at-issue. Tonhauser et al 2018 provided evidence for the GPP from utterances in which the relevant content C occurred in a polar question; in Exps 1, the at-issueness of the content was measured with the “asking whether” diagnostic. In this talk, which is based on joint work with Judith Degen (Stanford U), Taylor Mahler (Ohio State U), and Marie-Catherine de Marneffe (Ohio State U), I present the results of three experiments designed to further investigate the GPP. These experiments expanded on those reported in Tonhauser et al 2018 in that i) the content of the complement of both factive and non-factive predicates was investigated, ii) the clause-embedding predicates were embedded under negation or polar questions, and iii) additional measures for at-issueness reported in the literature were used. While the experiments provide further evidence for the GPP, they also point to the need to a) better understand measures of at-issueness, and b) consider the interaction between lexical semantics, projection, and at-issueness.
Shane Steinert-Threlkeld & Jakub Szymanik
Explaining two semantic universals of clause-embedding predicates
I will illustrate two broad approaches to explaining semantic universals—one based on learnability and one based on efficient communication—by looking at concrete examples from two domains of clause-embedding predicates. To the former, I will argue that responsive predicates which satisfy the property of veridical uniformity are easier to learn than those that do not. To the latter, I will present preliminary results (with Nathaniel Imel) suggesting that systems of modals that come closer to satisfying typological generalizations of Nauze and van der Klok are more optimized for efficient communication. The talk will conclude with some discussion of future directions and how to adjudicate between these styles of explanation.
Kristina Liefke
Meaning-driven combinatorial restrictions and ‘imagine *whether’
Unlike remember, imagine does not readily accept whether-complements. I argue that this difference cannot be explained through the factivity (or veridicality) of the embedding verb, as has been classically assumed. I propose, instead, to explain this difference (i) through the parasitic dependence of imagining on an underlying (veridical or counterfactual) experience and (ii) through the possibility of interpreting different constituents of the imagine-complement at different (viz. the imagination- or the experience-)alternatives (see Blumberg, 2018). I observe that whether-complements are only possible in scenarios in which the complement’s constituents are interpreted at the same alternatives (as is the case for ‘standard’ uses of remember, but only for few exceptional uses of imagine).
Nadine Theiler
Pragmatics intrudes into presupposition projection from wh-complements
How presuppositions project from matrix questions has been shown to depend on how these questions are used in discourse (Schwarz and Simonenko 2018, Theiler 2021): while canonical questions give rise to universal projection, certain kinds of non-canonical questions are associated with weaker projection patterns. In this talk, I will argue that we find a similar kind of pragmatic intrusion with projection from interrogative complements. How presuppositions project from these complements can depend, among other things, on what speaker and attitude holder know about the embedded question. I will propose to account for this by incorporating a bridge principle into the lexical semantics of the embedding predicates. With non-veridical predicates this principle governs projection to the modal base, while with veridical predicates it additionally gets imported into the matrix context via the veridicality of the modal base.
Jenny Doetjes
Interpreting French wh-phrases in situ
Tom Roberts
Parenthetically-embedded questions
Abstract
Donka Farkas & Floris Roelofsen
Encapsulated Questions
Embedded questions in English, exemplified in (1), normally do not involve subject-auxiliary inversion, unlike root questions, exemplified in (2).
(1) Paul wondered where he should go next.
(2) Where should he go next?
This paper is concerned with cases where questions have the form of root questions even though they are embedded in larger structures, as exemplified in (3). We call these encapsulated questions.
(3) Paul wonders where should he go next.
Building on work of McCloskey and Dayal, we propose a novel distinction between two types of encapsulated questions, ‘quasi-subordinate questions’ and ‘commented root questions’, and investigate the main empirical issues these constructions raise, namely (i) what types of questions may occur as encapsulated questions, (ii) what types of predicates allow encapsulation, and (iii) what is the interpretive effect encapsulation. We argue that the distinction between quasi-subordinate and commented root questions is relevant to all three of these issues. In our approach, encapsulated questions differ from plain subordinate ones in that, roughly speaking, they always raise an issue relative to some context. In the case of commented root questions, the issue is always raised relative to the current discourse context. In the case of quasi-subordinate questions, what context the issue is raised in depends on the semantic properties of the predicate.
Tanya Bondarenko
When clauses are Weak NPIs: polarity subjunctives in Russian
In this talk I investigate a class of verbs in Russian which take polarity subjunctives (Rivero 1971, Stowell 1993, Brugger & D’Angelo 1995, Giannakidou 1995, Giannakidou & Quer 1997, Quer 1998, Siegel 2009, Quer 2009, Giannakidou 2011, a.o.)—embedded subjunctive clauses whose acceptability depends on the properties of the environment. For example, in (1) we see that Russian pomnit’ ‘remember’ cannot take subjunctive clauses (morphologically expressed by the particle by that attaches to the complementizer) in an upward-entailing environment. However, when the embedding verb occurs under negation, in the scope of tol’ko ‘only’ or in a question, both indicative and subjunctive complements are possible, (2)-(4).
(1) Mitja pomnit, čto /*čto-by Anja kurila.
Mitya remembers COMP /COMP-SUBJ Anya smoked
‘Mitya remembers that Anya smoked.’
(2) Mitja ne pomnit, čto /čto-by Anja kurila.
Mitya NEG remembers COMP /COMP-SUBJ Anya smoked
‘Mitya doesn’t remember that Anya smoked.’
(3) Tol’ko Mitja pomnit, čto /čto-by Anja kurila.
only Mitya remembers COMP /COMP-SUBJ Anya smoked
‘Only Mitya remembers that Anya smoked.’
(4) Mitja pomnit, čto /čto-by Anja kurila?
Mitya remembers COMP /COMP-SUBJ Anya smoked
‘Does Mitya remember that Anya smoked?’
These data raise the question of how polarity subjunctives are licensed. Which verbs can they occur with and why? Why are they not possible with these verbs in upward-entailing environments? I propose that clauses can be existential quantifiers, and take scope (including exceptional scope). Subjunctive clauses are weak NPIs which have to be licensed in Strawson Entailment-Reversing environments. As other weak NPIs in Russian, they are not acceptable in non-monotone environments. Furthermore, I argue that whether a verb can take a polarity subjunctive or not is determined by the possible interpretations of CPs that combine with it. In my previous work I argued for two distinct meanings for Russian čto-clauses: they can either denote predicates of individuals with propositional content (Kratzer 2006, Moutlon 2015, Elliott 2017, a.o.), or predicates of exemplifying situations. I show that polarity subjunctives in Russian occur only with verbs that can take CPs that denote predicates of exemplifying situations. I argue that this restriction arises because the semantics of CPs that denote predicates of contentful individuals makes the environment non-monotone and thus prevents subjunctive from being licensed.
Craige Roberts
Meaning postulates in lexical pragmatics: Background vs. presupposition