One reason for interest in a formal theory of linguistic context is the possibility of what Pagin and Westerståhl (2010) call general compositionality. Yet the most sophisticated formal treatment of linguistic contexts currently available, due to Pagin and Westerståhl, both violates compelling intuitions about linguistic contexts and, as a result, does not recognize the discourse contexts of the dynamic Tarskian semantics due to Pickel and Rabern (2016). Thus general compositionality is undefined for any semantics using such contexts. In this paper, I present an alternative treatment of linguistic contexts and general compositionality adequate for understanding the contexts of Pickel and Rabern’s dynamic Tarskian semantics. Further justification for the present approach to compositionality and the semantics for variables is pursued elsewhere.
In this paper, we focus on the cross-linguistic incompatibility between the temporal adverbial ⌜then⌝ and the present tense. We show that the pattern holds in both matrix and embedded environments, across an array of languages (including Modern Greek, Modern Hebrew, Russian and Japanese) each behaving differently with respect to tense shift. To account for this incompatibility, we propose that the interpretation of tenses is sensitive to a perspective parameter. Tense shift occurs thanks to a perspective-shifting operator, similar to operator-based analyses of indexical shift (Anand and Nevins in Semantics and Linguistic Theory 14: 20, 2004; Anand in Ph.D. thesis Massachusetts Institute of Technology Cambridge, 2006; Deal in A Theory of Indexical Shift: Meaning, Grammar, and Crosslinguistic Variation, 2020). The main advantage of this account over alternative theories of tense shift, such as relative tense (Ogihara in Tense, Attitudes, and Scope. Vol. 58 of Studies in Linguistics and Philosophy, Springer, Dordrecht, 1996; Ogihara in Journal of Semantics 39: 581–616, 2022) or tense deletion (Ogihara and Sharvit in Embedded Tenses, Oxford University Press, 2012), is that we are able to distinguish between shifted and deleted tense. This is needed to explain the incompatibility of ⌜then⌝ with the shifted present but not with deleted past. Finally, we argue that the newly introduced perspective parameter needs to be separate from the speech context and the evaluation index.
Holton (2017) draws on the observation that no natural language has a contrafactive to argue that the object of knowledge is a fact. I fill in Holton’s argument sketch by asking how the truth and falsity presuppositions of, respectively, know and a contrafactive could arise. I defuse previous objections to his argument sketch, but, in the final section, suggest that Holton’s argument fails because presupposition triggering requires less than his argument assumes.
We propose a novel semantic analysis of the verb to know that takes the idea of question orientation of interrogative-embedding predicates to the limit. The new analysis conceptualizes knowledge as a judgment about an ability to answer questions and not as a sort of belief. This way the fact that the argument of to know is a question becomes crucial for the analysis of what knowledge is. The proposal naturally derives the different exhaustive interpretations of embedded questions from an interaction between the lexical semantics of the embedding predicates and a novel pragmatic principle. It also offers an explanation for several puzzles about question embedding under to know and seems easily extendable to other question-embedding verbs.
The goal of this paper is to develop a normative inferentialist account of open texture, inspired mainly by Robert Brandom (Making it explicit: Reasoning representing, and discursive commitment. Harvard University Press, 1994; Articulating reasons: An introduction to inferentialism. Harvard University Press, 2000a). At the level of metasemantics, open texture is explained by the fact that meaning is grounded in social normative practices that are themselves situated, plural, and contingent. While they are sufficiently stable to enable and sustain coordination, practice-based norms become unsettled in the face of disruptive or unexpected circumstances. At the level of semantics, open texture is explained by the fact that inferential norms are always incomplete and potentially incoherent. On the resulting view, open texture appears not just as a genuine phenomenon, but as a ubiquitous feature of concepts that pertains to the very nature of linguistic norms: there is a continuous spectrum between determinacy and indeterminacy, settled norms and unsettled norms. Open texture is visible only when norms become unsettled in the face of novelty and a creative resolution is needed in order to decide how to “go on”. In providing this account, I also reconstruct the historical and theoretical lineage connecting Waismann’s notion of open texture, Wittgenstein’s rule-following considerations, and Brandom’s normative inferentialism, showing that Brandom not only systematizes Wittgenstein’s view of meaning but also develops Waismann’s distinctive insights, such as the non-monotonicity of non-deductive rules of inference.
This paper examines Venn diagrams as a case study of visual representation. Venn diagrams have a clear formal structure and clearly defined correctness conditions. The question for this paper is whether these diagrams have a recursive syntax and corresponding compositional semantics by analogy with those developed for formal or natural languages. Existing research pointedly fails to offer a compositional semantics for Venn diagrams. This might suggest a fundamental cleavage between linguistic representation and the kind of visual representation exhibited by diagrams. Or, it might indicate a case of representation that recalcitrantly resists compositional analysis. This paper presents two approaches to the syntax of Venn diagrams: a region-first approach and a cell-first approach. It then develops a compositional semantics for each approach. The paper therefore helps to locate the similarities and differences more precisely between linguistic representation and the kind of representation exhibited by Venn diagrams. The two approaches to the syntax and semantics for Venn diagrams also suggest new avenues for interpreting other forms of visual representation.
In the following, I will explore the significance of the threat and challenge open texture can pose to analyticity. I take it that this challenge is very serious but has not been taken seriously enough so far. I will first discuss in some detail (Sect. 1) whether and, if yes, how there still can be necessary or sufficient conditions for some open textured concept or predicate expressing that concept. It turns out that the idea of synonymy does not apply in any straightforward, bivalent way to predicates expressing such a concept. The same holds for the idea of analyticity if it is explained via the idea of synonymy. All this does not mean that there is no tenable conception of analyticity without indeterminacy left (Sect.2). This discussion will also lead to a distinction between different levels of open texture: open texture can be iterated and spread across a system of concepts. The general conclusion is not that the analytic-synthetic distinction cannot be upheld but rather that, given how pervasive open texture is, there might be so few analytic truths that the distinction turns out to be much less significant than perhaps assumed. I end by raising some further questions for further inquiry. An appendix indicates further reasons, independent of open texture, to doubt that there are many analytic truths at all.
Presupposition triggers have been noted to vary in how easily they can be globally or locally accommodated. A look at previous classifications with respect to these two properties (e.g. Kripke 2009 for global accommodation, Abusch 2010 for local accommodation) shows that there is substantial overlap between triggers with respect to both types of accommodation difficulty (e.g. too is difficult both globally and locally, whereas stop is easy for both). This paper presents two sets of experiments comparing twelve triggers in English with respect to their global and local accommodation difficulty to test (i) whether the two accommodation types do in fact pattern together, and (ii) how prior classifications line up with the results. The results provide evidence for a unified approach of global and local accommodation, and are overall in line with prior classifications. Some of the unexpected parts of the results, including a contrast between the aspectual verbs stop and continue, and the role of context are discussed.
This paper offers a new theory of donkey anaphora that does not include any dynamic component. Even if the approach is not dynamic, it retains a key aspect of the dynamic tradition, namely the view that information states include not just factual information about the world, but also information about discourse referents, e.g., variables. It also makes crucial use of plural assignment functions (sets of standard assignments, cf. van der Berg 1996; Nouwen 2003; Brasoveanu in Linguist Philos 31: 129–209, 2008). Unlike dynamic approaches, sentences are evaluated as true or false relative to a pair (w, G), where w is a possible world and G is a plural assignment, with no reference to contexts or information states, and compositional semantics does not refer in any way to context update. In order to predict adequate meanings and felicity conditions, I combine two ingredients that have been used to account for presupposition projection, namely trivalence (Peters 1979, Beaver and Krahmer 2001) and Schlenker’s Transparency Condition (Schlenker 2007, 2008a). Two ideas play a crucial role in the proposal. First, a sentence such as ‘She _x came’ comes with the presupposition that the variable x is “valued” and denotes an atomic individual, which means that every atomic assignment in G maps x to the same atomic value. Second, I adopt Mandelkern’s witness condition (Mandelkern’s 2022): an existential statement such as ‘Someone _x came’ is undefined in (w, G) if it is classically true in w but G does not map x to a witness of the existential statement. Importantly, undefinedness is not equated with presupposition failure (e.g., even though ‘Someone _x came’ can be undefined, it is in fact never a presupposition failure). Rather, presupposition projection is governed by Schlenker’s Transparency Condition (Schlenker 2007, 2008a): the presupposition ‘x is valued and atomic’ should be redundant in the syntactic position in which ‘She _x came’ occurs. At the end of the paper, I discuss well-known ambiguities with donkey sentences (weak vs. strong, existential vs. universal readings) and show how they can be addressed in my system. The theory is presented here as a non-standard semantics for first-order logic, rather than a fragment of a natural language. Free variables are the counterparts of syntactically unbound pronouns, and existential quantifiers those of singular indefinites.
In this paper, I introduce the notion of insinuative reference and offer an account of the phenomenon it describes. In insinuative reference, speakers use demonstratives such that they have an innocuous possible referent and a less innocuous referent. The speaker intends for the hearer to recognise the less innocuous referent as a semantic value of the demonstrative in question. Yet, they also intend to preserve plausible deniability regarding that intention. Insinuative reference is often used to bring up politically taboo or otherwise offensive content with relative strangers and thus to undermine and avoid norms against such content. Analysing it is a form of ‘non-ideal philosophy of language’ as advocated by Beaver and Stanley. In the paper, I distinguish insinuative reference from insinuation proper (as described by Camp) and from over dogwhistling. I further describe how it poses a prima facie challenge to King’s “coordination account” of demonstratives. I show that this challenge can be met once we realise that speakers can have differing conceptions of reasonableness. In successful insinuative reference both interlocutors in fact accept that any reasonable hearer would recognise the speaker’s intention, yet this is not part of the common ground.
On one plausible story about how we rationally acquire reasons for action, an individual rationally acquires a new reason to ϕ only if there is something that individual learns about the world. It can also be said without much of a doubt that, in talking to each other, we sometimes rationally acquire reasons for action. This combination of claims is easy to square with declarative sentences, which we clearly use to provide each other with information when we make assertions. But it is much harder to square with imperative sentences, which are standardly taken not to be used in this way. How is it that in attempting to change someone’s behavior by giving them an instruction, we manage to acquaint them with reasons to act? Here, I offer a novel way out of this puzzle: instructions, I argue, are much more like assertions than we ordinarily take them to be, including in their being used to provide each other with information. Whereas assertions provide us with information about any arbitrary aspect of the world, instructions provide us with information specifically about what it is rational to conclude in the course of some practical inquiry, given our goals and the available evidence.
This paper is a reply to Mikkelsen and Hardt’s 2024 article analyzing same and different as involving an additive presuppositional element TOO akin to overt too. This hypothesis has several virtues: it provides a uniform analysis for both same and different under all of their (internal, external and comparative-like) readings; it explains several effects (poorly discussed so far) exhibited by same: parallelism and island effects, as well as analogies with too. But I show that this additive analysis faces several challenges questioning its explanatory power: the uniformity of the account rests on some stipulations, and the analogies between same/different and too are only partial. Furthermore, I show that Charnavel’s 2015 alternative account likening same/different to comparatives can derive the effects observed by Mikkelsen and Hardt without raising the problems induced by TOO incorporation.
In this paper, I argue that it is possible to understand Waismann’s notion of open texture through the paradigm of prototype theory stemming from cognitive linguistics. Indeed, prototype theory conceives concepts as functioning according to a prototype whose concrete application remains indeterminate and flexible. Prototype theory entails thus the possibility that cases that challenge the familiar use of a concept may be encountered in experience, which is precisely at stake with the notion of open texture. At the same time, prototype theory explains the creative dimension of human language that is also mentioned by Waismann: human language is creative precisely because its concepts are flexible and have subsequently the capacity of expressing new types of experience. This new perspective on the notion of open texture presents a double interest. First, it allows highlighting the innovative scope of this notion, since it shows that it has the potential to challenge traditional theories of concepts, which consider concepts as a set of characteristics. Second, it opens the possibility for an extension of the notion of open texture to most if not to all concepts. Several research works have been carried out in this direction. Reconsidering open texture through prototype theory offers a new theoretical tool for this purpose: proving that other, non-empirical concepts also function according to prototypes would indeed prove that open texture is an essential feature of most if not all concepts.
This paper argues that tense is used to mark evidential source in English. Specifically, nonfuture tenses mark the speaker’s evidence as causally downstream from the event being described. This amounts to a linguistic solution to a puzzle recently posed by Dilip Ninan (“Assertion, evidence and the future,” The Philosophical Review, 2022). The analysis is principally supported by parallel phenomena in evidential systems of Bulgarian and Korean.
We argue that per-phrases like two grams per liter exhibit a dual interpretive pattern. When such phrases saturate predicates of quantities in simplex dimensions, like weight or distance, they are interpreted compositionally within grammar; by contrast, when referring to quantities in quotient dimensions, like density or speed, they verbalize technical notation using a type of math speak. Akin to mixed quotations, such verbalizations get their meanings from extra-grammatical conventions. Supporting our analysis, per-phrases that refer to quotient dimensions, unlike their simplex counterparts, allow meaning-preserving substitution with expressions that unambiguously verbalize technical notation. Furthermore, these same expressions are opaque to syntactic operations. Our arguments challenge the view that per encodes quantity division compositionally, and underscore the distinction between meanings built in grammar and those licensed by extra-grammatical conventions.
Temporal in-adverbials lead a double life. Under one guise, they specify the durations of events; under another, they specify the durations of times throughout which certain events don’t take place. Each variety comes with its own seemingly idiosyncratic distributional restrictions. The distribution of the first class of expressions is restricted by the lexical aspect of VPs (Vendler, 1967; Dowty, 1979; Krifka, 1989, i.a.). The distribution of the second class is restricted by the polarity of sentences (Gajewski, 2005, 2007; Hoeksema, 2006; Iatridou Zeijlstra, 2017, 2021). I argue for a unified semantic analysis of both classes, which derives from one semantic principle their eclectic distribution: it must be possible for temporal in-adverbials to provide a maximally informative measure.
We propose the marginality scales account of the semantics of vague gradable adjectives, argue that this account is supported by Fara’s interest-relative conception of vagueness, and show how this fact sheds new light on Dinis and Jacinto’s (Erkenntnis 90:517–544, 2025a) nonstandard primitivism about vagueness—a central feature of the marginality scales account. Nonstandard primitivism relies on so-called theory. We also offer a significantly simplified version of this theory.
Any comprehensive semantics for number words, such as ‘two’, must explain why they can take on such a wide range of related meanings. Extant analyses offer promising explanations of a wide range of these meanings. On the other hand, they face a number of substantial, previously unrecognized theoretical and empirical challenges: they threaten incoherence, they identify numbers with the wrong sorts of entities, and they seemingly cannot explain the full range of meanings number words can take on. I argue that recognizing numbers as kinds affords a new type of polymorphic analysis, one that not only avoids these challenges, but also improves on extant explanations of polysemy, by expanding their empirical coverage.
This article argues for a logophoric analysis of Free Indirect Discourse (FID). FID is descriptively a hybrid between Direct Discourse (DD) and Indirect Discourse (ID). Recent studies largely agree on a DD-based analysis of FID by relying on bicontextual dependency (Schlenker in Mind Lang 19:279–304, 2004, Eckardt, The semantics of free indirect discourse: How texts allow us to mind-read and eavesdrop, Brill, Wiley, Leiden, 2014, i.a.) or mixed quotation (Maier in Mind Lang 30:345-373, 2015, i.a.). Instead, the article defends an ID-based, logophoric analysis of FID on the basis of overlooked properties of FID such as (anti)licensing of (anti)logophoric elements and recursive embedding of FID, which strengthen some previously discussed arguments such as de se and de te readings or sequence of tense phenomena (see Sharvit in Linguist Philos 31:353–395, 2008); the new observation that time and location adverbials are in fact not systematically indexicals anchored to the protagonist (but can be anaphoric or anchored to the speaker) further supports ID-based against DD-based analyses. The hypothesis that FID is outscoped by a logophoric operator not only derives the mixed properties of FID, but also treats FID as a case of an independently motivated linguistic class—the class of logophoric contexts.
According to Chomsky and followers, natural language is a computational system that generates syntactic structures that are counterfunctional with respect to communication. Consequently, language is more appropriately considered as being “designed” for thought rather than communication. In this paper, we argue that, while natural language, understood as an internal computational system along standard generative lines, is recruited for distinctive human thinking, such recruitment also requires, and is strongly influenced by, a process we dub lexical externalisation. We first show that there is good reason to believe that the atomic items that go to form words essentially include phonological information of externalised words. After that, we explore what externalisation implies with respect to the relation between lexical items and concepts. We suggest that externalisation has profound effects on what concepts we think with, as well as how concepts relate to lexical items (i.e., how thinking processes ultimately relate to linguistic computations).
