Approaches to negativity tags, like English neither-tags, ‘not even‘-tags, and why not, differ in whether they treat the antecedent as a clause with clausal negation (e.g., Klima 1964; Kramer and Rawlins 2009; Brasoveanu et al. 2013) or, more generally, one introducing a salient counterfactual proposition into the discourse (Hofmann 2022, 2023). This paper provides experimental evidence for the latter view, showing that negativity-tags are rated as natural (compared to an affirmative baseline) when their antecedent is introduced without clausal negation, but in an anti-veridical embedding (e.g., It’s a lie), under neg-raising (e.g., I don’t get the feeling), or with ‘lexical indicators of irony’ (Horn 2016, e.g., as if ).

The XPrag slides present Exp. 1 about counterfactual content, and Exp. 2 about at-issueness; The SuB slides present Exp. 1, as well as recipe for making negativity-tags, and a formal semantic analysis of negative additive tags.

This work develops a dynamic analysis of presupposition projection based on similarity relations between local and global contexts, drawing on accounts of counterfactual conditionals (Lewis 1973, Stalnaker 1978). The proposal addresses issues with presuppositions triggered in counterfactual contexts, where existing local-context based accounts (Heim 1983, 1992, Schlenker 2009) make faulty predictions when considering counterfactual contexts that are logically distinct from the global context. This work addresses this issue, deriving projection by requiring the global context to be similar to the local context. It also addresses the triggering problem by treating presuppositions as entailments (Abrusán 2011) and deriving projection by calculating similarity over partitions of the context set determined by the question under discussion (Simons et al. 2017). The analysis is implemented in a propositional version of CDRT, using propositional variables to model local and global context sets (Brasoveanu 2006, AnderBois et al. 2015, Hofmann 2022, 2025).

At-issueness is a key concept in theoretical semantics/pragmatics, but there is no consensus about how it is defined or diagnosed (e.g., Tonhauser 2012; Snider 2017b; 2018; Tonhauser et al. 2018; Koev 2018; Faller 2019; Korotkova 2020). We present experimental data investigating whether four widely used diagnostics for at-issueness yield consistent results. Our findings reveal significant differences across diagnostics, indicating they are not interchangeable, and that the speech-act in which the diagnosed content is embedded matters: We get a clearer picture when assessing the not-at issue content of different expressions in questions rather than assertions.

This paper presents an intensional account of anaphoric accessibility, providing a unified analysis of anaphora with antecedents under negation and non-veridical operators. These include cases with double negation, bathroom-disjunctions, and modal subordination, which have previously received disparate analyses. Building on analyses of modal subordination (Stone 1999; Brasoveanu 2006), the paper introduces a flat-update dynamic semantics, where expressions with the potential to introduce a discourse referent do so globally, regardless of their embedding context. Constraints on anaphora are derived based on discourse consistency and a pronominal existential inference interpreted relative to possible worlds in the local context. The analysis uses intensional representations of discourse referents, which store information about their embedding context, and its relationship to speaker commitments, explaining constraints imposed by negation and other non-veridical operators in terms of presuppositions about discourse referents.

This chapter explores the semantics of anaphora by highlighting analytical challenges posed by anaphora with unspecific antecedents and reviewing solutions proposed in semantic theory. The main focus is how pronouns establish reference in hypothetical and counterfactual discourse, contrasting descriptive and dynamic approaches in formal semantics. It provides a detailed discussion of dynamic approaches to the challenges in tracking referents across hypothetical and counterfactual discourse contexts. The discussion also connects to philosophical notions of reference and includes a review of psycholinguistic experiments investigating the cognitive mechanisms involved in interpreting anaphora. The chapter is intended to be useful to advanced students of linguistic semantics and researchers interested in reading about the formal semantics of anaphora, or its intersection to theories of language processing.

Under the ‘family of sentences’ diagnostic for projection, the projection of content is investigated by embedding the expression that contributes the content in the scope of negation, polar questions, epistemic possibility modals, and conditional antecedents. This paper reports on the results of a set of experiments designed to investigate whether there is variation in the projection of content from under these four types of entailment-canceling operators. The contents investigated are the contents of the complements of 20 English clause-embedding predicates. The results of the experiments suggest (i) that the by-operator variation is small when aggregating over the 20 contents, but (ii) that the effect of operator differs between the clause-embedding predicates. The results of these experiments also extend a result of Degen and Tonhauser 2022, that projection ratings in polar questions do not categorically distinguish factive from non-factive predicates, to cases with negation, the epistemic possibility modal perhaps, and conditional antecedents. The observed by-predicate and by-operator variation is not captured by existing theoretical accounts of projection (e.g., Heim 1983, van der Sandt 1992, Abrusán 2011, Schlenker 2021). Our results suggest that an empirically adequate projection analysis must consider interactions between predicates and operators.

This paper addresses the anaphoric polarity sensitivity of negativity-tags, challenging the idea that they are only licensed by sentential negation (e.g., Klima 1964; Kramer and Rawlins 2009; Farkas and Bruce 2010; Brasoveanu et al. 2013, 2014; Roelofsen and Farkas 2015), arguing instead that they are sensitive to counterfactual propositional content in discourse. This is supported by data showing that negativity-tags are licensed without overt negation and their acceptability is influenced by contextual factors. The emerging notion of discourse-polarity has theoretical implications: The discourse-effect of negation is tied to its anti-veridical semantics, and characterizing negative antecedents requires a discourse-level representation that integrates information from both semantic representations and pragmatic inferences.

Certain propositional anaphora, like the Polarity Particles and Polar Additives discussed in this paper, are sensitive to the polarity of their antecedent clause. The paper establishes that discourse polarity—the polarity of the antecedent clause for the purposes of licensing subsequent anaphora—is influenced by complex factors, some of them syntactic, semantic, and pragmatic in nature. The hyperintensional dynamic framework presented here gives a formal foundation to a distinction of polarity for propositional discourse referents and captures some of the central generalizations. It uses discourse referents for hyperintensional propositions, providing a level of representation that connects information from the discourse context, the proposition’s semantic content, and the information about the polarity of the antecedent clause. Therefore, it constitutes a step towards an analysis capturing the heterogeneous factors influencing discourse polarity.

My dissertation (advised by Adrian Brasoveanu) investigated the interaction of anaphora and negation. The work provides arguments and evidence for theoretically understanding the constraints that negation places on anaphora in terms of speaker commitments and veridicality, providing a commitment-based account of discourse-polarity and anaphoric polarity-sensitivity by appealing to veridicality distinctions.

On these terms, a formal semantic account of the interaction of anaphora and negation is developed. It is an intensional dynamic account based on Compositional DRT and flat-update systems, where negation and other non-veridical operators are treated as externally dynamic. I argue that we need to conceptualize discourse referents as an intensional and epistemic representation to capture the relevant interactions.

This provides a unified account of several related issues that have received disparate analyses in the previous literature: anaphora to indefinites under negation (double-negation, bathroom-disjunctions, discourse subordination, and cross-speaker cases), the interaction of propositional anaphora and negation, polarity-sensitive negativity-tags, and the question what counts as a negative sentence/utterance for the purposes of anaphora in discourse.

This work argues that the pre-nominal particle mí in combination with an NP in San Martín Peras Mixtec (Oto-Manguean, Oaxaca, Mexico) could be characterized as a definiteness marker. Further, mí occurs with pronouns, where it often receives a reflexive translation. Although ‘mí + Pronoun’ usually requires a local antecedent, many occurrences couldn’t be characterized as reflexive, logophoric, or bound. We are currently investigating the question of what precisely the locality-constraints on its anaphoric behavior are.

This paper extends the analysis of modal subordination presented in Stone (1999); Stone and Hardt (1999) to other cases of non-veridically introduced discourse referents (drefs) with the goal of understanding the circumstances under which a dref is available for subsequent anaphoric reference. The central claim is that a dref introduced under negation can be the antecedent for a pronoun, only if the interpretation does not require that the assertion of its non-existence and the existence presupposition of the pronoun be true wrt the same set of worlds. The analysis is based on relativizing individual drefs to the worlds in which their referent exists (Stone (1999); Stone and Hardt (1999)), and sentential operators introducing propositional drefs that provide a local context for the interpretation of their prejacent (Karttunen (1973); Heim (1983)), and it is framed in intensional CDRT (Muskens (1996); Brasoveanu (2007, 2010)). This allows for a greater empirical coverage than Krahmer and Muskens’s (1995) account of drefs under negation and disjunction.

A known difference between Negative Indefinites (NIs) such as English nobody, and Negative Polarity Items (NPIs) such as anybody is that only the former may occur as fragments (Giannakidou 2000; Merchant 2004). This holds regardless of the polarity of the wh-Q antecedent.

1. a. Who did Mary see?
   b. Nobody. / *Anybody.
   ‘There is nobody that she saw.’

2. a. Who didn’t Mary see?
   b. Nobody. / *Anybody.
   ‘There is nobody that she didn’t see’ = ‘She saw everyone.’

This work addresses fragments involving NIs (NI fragments), across a variety of antecedent types. For wh-question (wh-Q) antecedents, the interpretation of the response varies with the polarity of the question. The work addresses how these these interpretations arise and what they tell us about the structure of fragments, the composition of NIs, and the parallelism that holds between an elided TP and its antecedent?

Reduced interrogative clauses with a causal wh-word and a negative marker (like English ‘Why not’) can have (at least) two kinds of readings:

1. Factive reading:
   A: Mary didn’t sleep.
   B: (Really?) Why not?

2. Modal reading:
   A: Let’s sleep.
   B: (Sure,) Why not?

The paper outlines different properties of the two readings in terms of their interpretation and anaphoric behavior. The factive reading receives its interpretation in combination with a salient negative proposition in the context. (1) can be paraphrased as Why didn’t Mary sleep?. The modal reading receives a modal interpretation in relation to a salient antecedent in the context. The literal content of (2) can be paraphrased as Why shouldn’t we sleep? The modal reading is generally accompanied by a rhetorical question inference, like There is no reason we shouldn’t sleep in (2). The paper outlines properties of the two readings, which differ in their licensing conditions, anaphoric behavior, and cross-linguistic realization. It also provides an analysis of the interaction of the factive reading and negation based on a syntactic dependency between the negative remnant (why not) and negation in the ellipsis site.

This paper reassesses the empirical basis for Kratzer and Selkirk’s (2007) phase-based account of prosodic phrasing in German, which argues that prosodic structure follows syntactic phase-based spell-out. It examines two key claims: (1) the contrast in main stress placement between stative and eventive intransitive clauses, attributed to their syntactic differences, and (2) the claim that VP-internal PPs are never stressed in neutral utterances, with stress always falling on the c-commanding accusative argument. I argue that the intransitive stress contrast follows from Diesing’s (1990) syntactic analysis and the nuclear stress rule, while the VP-internal PP stress pattern is better explained by information structure, as these PPs can be interpreted as given. The findings suggest that syntax and information structure, rather than phase-based prosody, better account for German stress patterns.

I wrote my BA thesis (supervised by Daniel Altshuler) on the semantics and pragmatics of performative utterances, discussing how the self-verification in performative utterances could be derived from the assertion of a self-referential semantic representation. At the interface of semantics, syntax, and pragmatics, this project combines questions about event reference, event participants, and individuation of events with questions about performative utterances.