McGee (1985) argued that modus ponens was invalid for the natural language conditional 'If. . . then. . . '. Many subsequent responses have argued that, while McGee's examples show that modus ponens fails to preserve truth, they do not show that modus ponens fails to preserve rational full acceptance, and thus modus ponens may still be valid in the latter informational sense. I show that when we turn our attention from indicative conditionals (the focus of most of the literature to date) to subjunctive conditionals, we find that modus ponens does not preserve either truth or rational full acceptance, and thus is not valid in either sense. In concluding I briefly consider how we can account for these facts.
We argue that definite noun phrases give rise to uniqueness inferences characterized by a pattern we call definiteness projection.
Definiteness projection says that the uniqueness inference of a definite projects out unless there is an indefinite antecedent in a position that filters presuppositions.
We argue that definiteness projection poses a serious puzzle for e-type theories of (in)definites; on such theories,
indefinites should filter existence presuppositions but not uniqueness presuppositions.
We argue that definiteness projection also poses challenges for dynamic approaches,
which have trouble generating uniqueness inferences and predicting some filtering behavior,
though unlike the challenge for e-type theories, these challenges have mostly been noted in the literature,
albeit in a piecemeal way. Our central aim, however, is not to argue for or against a particular view, but rather to formulate and motivate a generalization about definiteness which any adequate theory must account for.
I show that standard dynamic approaches to the semantics of epistemic modals invalidate the classical laws of excluded middle and non-contradiction, as well as the law of 'epistemic non-contradiction'.
I argue that these heretofore unnoticed facts pose a serious challenge.
I discuss what I call practical Moore sentences:
sentences like 'You must close your door, but I don't know whether you will',
which combine a command with an admission that the speaker does not know whether she will be obeyed.
Practical Moore sentences are infelicitous.
I argue that this infelicity is quite surprising:
intuitively, there should be nothing wrong with giving someone an order while acknowledging that you do not know whether it will obeyed.
I argue that this infelicity points towards a striking fact about human psychology:
when giving an order, we must act as if we believe we will be obeyed.
Sentences containing negated 'neg-raising' predicates like 'think' and 'want' license surprisingly strong inferences. These sentences
are generally interpreted in the same way as when negation takes scope below the embedding predicate.
For instance, 'James doesn't think that Marie was promoted' is generally interpreted in the same way as 'James thinks that Marie wasn't promoted'.
Neg-raising has also been linked to the licensing of strong Negative Polarity Items (NPIs), a subclass of NPIs like 'in years', as in
'James doesn't think that Marie has seen her mother in years'.
There are two main approaches to neg-raising in the literature.
The first approach is syntactic, and holds that, in neg-raising sentences, negation, while appearing in the main clause at the level of surface structure,
is actually in the embedded clause at the syntactic level that feeds into semantic interpretation.
The second is a semantic/pragmatic approach, in which neg-raising arises as an inference due to an excluded middle presupposition or implicature.
In this squib, we present data involving attitude predicates like 'appreciate', 'be glad', 'be happy', and 'like'.
These both give rise to neg-raising interpretations and presuppose the truth of their sentential complements; intriguingly, they do not license strong NPIs.
We argue that accounting for the neg-raising interpretation of the negations of these predicates, and the fact that they fail to license strong NPIs,
is a challenge for existing semantic and syntactic approaches to neg-raising, but that
the semantic approach, together with the theory of strong NPIs due to Gajewski (2011) and Chierchia (2013), can more easily accommodate these cases.
Import-Export says that a conditional with the form 'If p, if q, r' is always equivalent to the corresponding conditional 'If p and q, r'.
I argue that Import-Export does not sit well with a classical approach to conjunction: given some plausible and widely accepted principles about conditionals, Import-Export together with classical conjunction leads to absurd consequences.
My main goal is to draw out these surprising connections.
In concluding I argue that the right response is to reject Import-Export and adopt instead a limited version which better fits natural language data; accounts for all
the intuitions that motivate Import-Export in the first place; and fits better with a
In a brief discussion of epistemic modals, Wittgenstein (1953) warns against ‘regard[ing] a hesitant assertion as an assertion of hesitancy’.
A modal claim like ‘It might be raining’, the thought goes, should not be regarded as an assertion of the speaker's uncertainty as to whether or not it is raining,
but rather as something quite different in kind:
a proposal to treat the possibility of rain as live.
Wittgenstein’s admonition has, in recent years, been at the heart of arguments that, in order to make sense of the dynamics of epistemic modal claims,
we must reject the contextualist framework for analyzing communication—a framework on which assertions of epistemic modal claims, like all assertions, convey information.
In this paper, I argue that, on the contrary, taking Wittgenstein's admonition seriously does not require abandoning the contextualist framework:
we can capture the fundamental dynamics of epistemic modality within the contextualist framework, provided we take the assertoric content of unembedded epistemic modal claims
to be determined by the prospective common attitudes of the conversants in question.
Is the mechanism behind presupposition projection and filtering fundamentally asymmetric or symmetric?
This is a foundational question for the theory of presupposition which has been at the centre of attention in the literature recently (Schlenker 2008b, 2009; Rothschild 2011/2015 a.o.).
It also bears on broader issues concerning the source of asymmetries observed in natural language:
are these simply rooted in superficial asymmetries of language use (since language use unfolds in time,
which we experience as fundamentally asymmetric) or are they, at least in part, directly encoded in linguistic knowledge and representations?
In this paper we aim to make progress on these questions by exploring presupposition projection across conjunction,
which has typically been taken as a central piece of evidence that pre- supposition filtering is asymmetric in general.
As a number of authors have recently pointed out, however,
the evidence which has typically been used to support this conclusion is muddied by independent issues concerning judgments of redundancy,
and additional concerns arise with regards to the possibility of local accommodation. We report on a series of experiments, building on previous work by Chemla & Schlenker (2012);
Schwarz (2015), using inference and acceptability tasks, which aim to control for both of these potential confounds. In our results, we find strong evidence for asymmetric
left-to-right filtering across conjunctions, but no evidence for right-to-left filtering—even when right-to-left filtering would, if available, rescue an otherwise unacceptable sentence.
These results suggest that presupposition filtering across conjunction is indeed asymmetric, contra suggestions in the recent literature (Schlenker 2008a, 2009 a.o.)
and paves the way for the investigation of further questions about the nature of this asymmetry and presupposition projection more generally.
Our results also have broader methodological and theoretical implications: we find important differences in the verdicts of acceptability versus inference tasks in
testing for projected content, which has both methodological ramifications for the question of how to distinguish presupposed content, and theoretical repercussions
for understanding the nature of projection and presuppositions more generally.
I explore the logic of the conditional, using credence judgments to argue against Duality and in favor of Conditional Excluded Middle.
I then explore how to give a theory of the conditional which validates the latter and not the former, developing a variant on Kratzer (1981)'s restrictor theory,
as well as a proposal which combines Stalnaker (1968)'s theory of the conditional with the theory of epistemic modals I develop in Mandelkern 2019a.
I argue that the latter approach fits naturally with a conception of conditionals as referential devices which allow us to talk about particular worlds.
When embedding data are used to argue against semantic theory A and in favor of semantic theory B,
it is important to ask whether A could, after all, make sense of those data.
It is possible to ask that question on a case-by-case basis.
But suppose we could show that A can make sense of all the embedding data which B can possibly make sense of.
This would, in one fell swoop, undermine all arguments in favor of B over A on the basis of embedding data.
And, provided that the converse does not hold—that is, that A can make sense of strictly more embedding data than B can—it
would also show that there is a precise sense in which B is more constrained than A,
yielding a pro tanto simplicity-based consideration in favor of B.
In this paper I develop formal tools which allow us to make comparisons of this kind, which I call comparisons of potential expressive power.
I motivate the development of these tools by way of the recent debate about epistemic modals.
These tools show that several prominent revisionary theories which have been developed in response to facts about how epistemic modals embed are strictly less expressive
than the standard relational theory, in the sense that the relational theory can make sense of any embedding data involving epistemic modals which those theories
can make sense of, but not vice versa.
This necessitates a fundamental reorientation in how to think about the choice between these semantics for epistemic modals,
and yields a formal tool for comparing different semantics of fragments of natural language with broad applicability.
In concluding I consider in more detail the empirical picture concerning the behavior of epistemic modals in attitude contexts and what this
tells us about the theory of epistemic modals in light of the foregoing results.
There is a difference between the conditions in which one can felicitously use a 'must'-claim like (1-a) and those in which one can use the corresponding claim without the 'must', as in (1-b):
(1) a. It must be raining out. b. It is raining out.
It is difficult to pin down just what this difference amounts to.
And it is difficult to account for this difference, since assertions of 'Must p' and assertions of p alone seem to have the same basic goal:
namely, communicating that p is true.
In this paper I give a new account of the conversational role of 'must'.
I begin by arguing that a 'must'-claim is felicitous only if there is a shared argument for the proposition it embeds.
I then argue that this generalization, which I call Support, can explain the more familiar generalization that 'must'-claims are felicitous only if the speaker's evidence for them is in some sense indirect.
Finally, I propose a pragmatic derivation of Support as a manner implicature.
What does 'might' mean?
One hypothesis is that 'It might be raining' is essentially an avowal of ignorance like 'For all I know, it's raining'.
But it turns out these two constructions embed in different ways, in particular as parts of larger constructions like Wittgenstein's 'It might be raining and it's not' and Moore's 'It's raining and I don't know it', respectively.
A variety of approaches have been developed to account for those differences.
All approaches agree that both Moore sentences and Wittgenstein sentences are classically consistent.
In this paper I argue against this consensus.
I adduce a variety of new data which I argue can best be accounted for if we treat Wittgenstein sentences as being classically inconsistent.
This creates a puzzle, since there is decisive reason to think that 'Might p' is classically consistent with 'Not p'.
How can it also be that 'Might p and not p' and 'Not p and might p' are classically inconsistent?
To make sense of this situation, I propose a new theory of epistemic modals and their interaction with embedding operators.
This account makes sense of the subtle embedding behavior of epistemic modals, shedding new light on their meaning and, more broadly, the dynamics of information in natural language.
When do we judge that someone was forced to do what they did?
One relatively well-established finding is that subjects tend to judge that agents were not forced to do actions when those actions violate norms.
A surprising discovery of Young and Phillips 2011 is that this effect seems to disappear when we frame the relevant 'force'-claim in the active rather than passive voice
('X forced Y to φ' vs. 'Y was forced to φ by X').
Young and Phillips found a similar contrast when the scenario itself shifts attention from Y (the forcee) to X (the forcer).
We propose that these effects can be (at least partly) explained by way of the role of attention in the setting of quantifier domains which in turn play a role in the evaluation of
We argue for this hypothesis by way of an experiment which shows that sequences of active vs. passive 'force'-claims display the characteristic "stickiness" of quantifier domain expansion,
using a paradigm which we argue provides a useful general paradigm for testing quantifier domain hypotheses.
Finally, we sketch a semantics for 'force' which we argue is suitable for capturing these effects.
Richard Bradley offers a quick and convincing argument that no Boolean semantic theory for conditionals can validate an apparently valid principle concerning the relationship between credences and conditionals.
We argue that Bradley's crucial principle, Preservation, is in fact invalid; its appeal arises from the validity of a nearby, but distinct, principle, which we call Local Preservation, and which Boolean semantic theories can non-trivially validate.
A part of Stalnaker (1968)'s influential theory of conditionals has been neglected, namely the role for an accessibility relation between worlds. I argue that the accessibility relation does not play the role intended for it in the theory as stated, and propose a minimal revision which solves the problem, and brings the theory in line with the formulation in Stalnaker & Thomason 1970.
Inquiry into the meaning of 'logical' terms in natural language ('and', 'or', 'not', 'if') has generally proceeded along two dimensions.
On the one hand, semantic theories aim to predict native speaker intuitions about the natural language sentences involving those logical terms.
On the other hand, logical theories explore the formal properties of the translations of those terms into some logical system.
Sometimes, these two lines of inquiry appear to be in tension: for instance, our best logical investigation into conditional connectives may show that there is no conditional operator that has all the properties native speaker intuitions suggest 'if' has.
Indicative conditionals have famously been the source of one such tension, ever since the triviality proofs of both Lewis 1976 and Gibbard 1981, whose conclusions are extremely implausible by the lights of ordinary judgments about natural language indicative conditionals.
In a recent series of papers, Branden Fitelson has strengthened both triviality results (Fitelson 2013, 2015, 2016), revealing a common culprit: a logical schema known as Import-Export.
Fitelson's results focus the tension between the logical results and ordinary judgments, since Import-Export seems to be supported by intuitions about natural language.
In this paper, we argue that the intuitions which have been taken to support Import-Export are really evidence for a closely related, but subtly different, principle.
We show that the two principles are independent by showing how, on a standard assumption about the conditional operator in the formal language in which Import-Export is stated, many existing theories of indicative conditionals validate one, but not the other.
Moreover, we argue that once we clearly distinguish these principles, we can use propositional anaphora to show that Import-Export is, given this assumption about the formal conditional operator, in fact not valid for natural language indicative conditionals.
This is an important advance, since it gives us a principled and independently motivated way of rejecting a crucial premise in many triviality results, while still making sense of the speaker intuitions which appeared to motivate that premise.
We suggest that this strategy has broad application and an important lesson: in theorizing about the logic of natural language, we must pay careful attention to the translation between the formal languages in which logical results are typically proved, and natural languages which are the subject matter of semantic theory.
Two recent and influential papers, van Rooij 2007 and Lassiter 2012, propose solutions to the proviso problem which make central use of related notions of independence, qualitative in the first case, probabilistic in the second.
We argue here that, if the solutions are to work, they must incorporate an implicit assumption about presupposition accommodation, namely that accommodation does not interfere with existing qualitative or probabilistic independencies.
We show, however, that this assumption is implausible, as updating beliefs with conditional information does not in general preserve independencies. We conclude that van Rooij and Lassiter's approach does not succeed in resolving the proviso problem.
(2) sounds entirely natural, whereas (1) sounds quite strange.
This contrast is puzzling, because (1) and (2) have the same structure at a certain level of logical abstraction.
We argue that existing theories of informational oddness do not distinguish between (1) and (2).
We do not have an account of the divergence in judgments about the two, but we think this is a fascinating puzzle which we pose here in the hope others will be able to solve it.
I discuss what I call practical Moore sentences: sentences which combine a command with an admission that the speaker does not know whether she will be obeyed. Practical Moore sentences are infelicitous.
I argue that this infelicity is surprising, and can be best explained by adopting a striking hypothesis about human psychology: when giving an order, we must act as if we believe we will be obeyed.
2017. Semantics and Linguistic Theory. 27:504-524.
[abstract] [local file] This is a proceedings paper further developed in ‘We've discovered projection is asymmetric (and it is!)’
Is presupposition projection a fundamentally asymmetric or symmetric matter?
That is, when processing presuppositions, can we take into account only material that precedes (either in time, linear order, or hierarchical order) a presupposition trigger, or can we also access material that follows the trigger?
This question is fundamental for the theory of presupposition as well as broader questions about whether observed asymmetries in language stem just from the superficial asymmetries of language use—language use unfolds in time, which we experience as fundamentally asymmetric—or from asymmetries encoded in linguistic knowledge.
In this paper we aim to make progress on these questions by exploring presupposition projection across conjunction, which has typically been taken as a central piece of evidence that presupposition projection is asymmetric.
As recently argued in the literature, however, it is much less clear than is commonly accepted once we take into account independent issues about redundancy (Rothschild 2008, 2011, Chemla and Schlenker 2011, a.o.).
When we look at cases that control for such issues, intuitive judgments about asymmetry become much less clear.
Similarly, however, intuitive judgments about potential opposite right-to-left filtering are equally difficult to disentangle from independent issues about presupposition suspension.
The question as to whether presupposition filtering in conjunction is asymmetric once we control for other factors is therefore entirely open.
Building on previous work by Chemla and Schlenker (2011), we approach the question experimentally by using an inference tasks controlling for redundancy and presupposition suspension.
In our results, we find strong evidence for left-to-right asymmetric filtration of presuppositions across conjunctions, but no evidence for right-to-left filtering.
We propose a new analysis of a class of modals which we call agentive modals: ability modals and their duals, compulsion modals. After criticizing existing approaches - the existential quantificational analysis, the universal quantificational analysis, and the conditional analysis - we lay out a new account that builds on both the existential and conditional analyses. On our account, a sentence like ‘John can swim across the river’ says that there is some practically available action (in a sense we make precise) which is such that, if John tries to do it, he swims across the river. We argue that this approach avoids the problems faced by existing accounts of agentive modality, and show how it can be extended to an account of generic agentive modal claims. The upshot is a new vantage point on the role of agentive modal ascriptions in practical discourse: ability ascriptions serve as a kind of hypothetical guarantee, on our account, and compulsion ascriptions as a kind of non-hypothetical guarantee.
We use antecedent-final conditionals to formulate two problems for parsing-based theories of presupposition projection and triviality of the kind given in Schlenker (2009). We show that in these cases parsing-based theories predict filtering of presuppositions where there is in fact projection, and triviality judgments for sentences which are in fact felicitous. More concretely, these theories predict that presuppositions triggered in the antecedent of antecedent-final conditionals will be filtered (i.e. will not project) if the negation of the consequent entails the presupposition. But this appears wrong: (1) intuitively presupposes that John is in France, contrary to this prediction.
(1) John isn't in Paris, if he regrets being in France.
Likewise, parsing based approaches to triviality predict that material entailed by the negation of the consequent will be redundant in the antecedent of the consequent; but (2) is intuitively felicitous, contrary to these predictions.
(2) John isn't in Paris, if he's in France and Mary is with him.
Importantly, given that the trigger appears in sentence-final position, both incremental left-to-right and symmetric versions of such theories make the same predictions.
These data constitute a challenge to the idea that presupposition projection and triviality should be computed on the basis of parsing. This issue is important because it relates to the more general question as to whether presupposition and triviality calculation should be thought of as a pragmatic post-compositional phenomenon or as part of compositional semantics, as in the more traditional dynamic approaches. We discuss a solution which allows us to maintain the parsing based pragmatic approach; it is based on an analysis of conditionals which incorporates a presupposition that their antecedent is compatible with the context, together with a modification to Schlenker (2009)'s algorithm for calculating local contexts so that it takes into account presupposed material. As we discuss, this solution works within a framework broadly similar to (though in some interesting ways different from) that of Schlenker's, but it doesn't extend in an obvious way to other parsing-based accounts, e.g. trivalent ones.
We use antecedent-final conditionals to formulate a challenge to parsing-based theories of local contexts, and associated theories of presupposition projection and triviality, like the one given in Schlenker 2009.
We show that a theory like Schlenker's predicts that the local context for the antecedent of an antecedent-final conditional will entail the negation of the conditional's consequent.
It thus predicts that presuppositions triggered in the antecedent of antecedent-final conditionals will be filtered if the negation of the consequent entails the presupposition.
But this is wrong: 'John isn't in Paris, if he regrets being in France' intuitively presupposes that John is in France, contrary to this prediction.
Likewise, parsing-based approaches to triviality predict that material entailed by the negation of the consequent will be felt to be redundant in the antecedent of the conditional.
But this is wrong: 'John isn't in Paris, if he's in France and Mary is with him' is intuitively felicitous, contrary to this prediction.
Importantly, given that the material in question appears in sentence-final position in antecedent-final conditionals, both incremental (left-to-right) and symmetric versions of parsing-based theories of local contexts make the same problematic predictions here.
In Mandelkern and Romoli 2017, we discuss one solution to this problem, given within a broadly parsing-based pragmatic approach.
In this paper, we explore an alternate direction: incorporating attention to hierarchical structure into the calculation of local contexts.
We sketch several possible implementations and point to some of the possibilities and challenges for a hierarchical approach to local contexts.
There is a difference between the conditions in which one can felicitously use a ‘must’-claim like (1), versus those in which one can use the corresponding claim without the ‘must’. But it is difficult to pin down just what this difference amounts to. And it is even harder to account for this difference, since assertions of ‘Must p’ and assertions of p alone seem to have the same basic effect: namely, sharing the information that p is true. In this paper I take on these two subtle and long-standing puzzles, puzzles which get to the heart of questions about the meaning of epistemic modals, the norms that govern assertions, and the way we process and organize information. I begin by arguing on the basis of novel cases and experimental data for the empirical generalization that a ‘must’-claim is felicitous only if there is a shared argument for the proposition it embeds. I then argue that this generalization, which I call Support, can explain the more familiar generalization that ‘must’-claims are felicitous only if the speaker's evidence for them is in some sense indirect. Then I propose a pragmatic derivation of Support as a manner implicature. I close by showing how my account can be extended to explain cases in which a non-modal claim is unacceptable but a ‘must’-claim is felicitous.
I propose a new theory of semantic presupposition, which I call dissatisfaction theory. I first briefly review a cluster of problems - known collectively as the proviso problem - for most extant theories of presupposition, arguing that the main pragmatic response to them faces a serious challenge. I avoid these problems by adopting two changes in perspective on presupposition. First, I propose a theory of projection according to which presuppositions project unless they are locally entailed. Second, I reject the standard assumption that presuppositions are contents which must be entailed by the input context; instead, I propose that presuppositions are contents which are marked as backgrounded. I show that, together, these commitments allow us to avoid the proviso problem altogether, and generally make plausible predictions about presupposition projection out of connectives and attitude predicates. I close by sketching a two-dimensional implementation of my theory which allows us to make further welcome predictions about attitude predicates and quantifiers.
The Proviso Problem is the discrepancy between the predictions of nearly every major theory of semantic presupposition about what is semantically presupposed by conditionals, disjunctions, and conjunctions, versus observations about what speakers of certain sentences are felt to be presupposing. I argue that the Proviso Problem is a more serious problem than has been recognized in much of the current literature. After briefly describing the problem and two standard responses to it, I give a number of examples which, I argue, show that those responses are inadequate. I conclude by briefly exploring alternate approaches to presupposition that avoid this problem.
We propose a new analysis of ability modals. After briefly criticizing extant approaches, we turn our attention to the venerable but vexed conditional analysis of ability ascriptions. We give an account that builds on the conditional analysis, but avoids its weaknesses by incorporating a layer of quantification over a contextually supplied set of actions.
The meaning of definite descriptions (like 'the King of France', 'the girl', etc.) has been a central topic in philosophy and linguistics for the past century. Indefinites ('Something is on the floor', 'A child sat down', etc.) have been relatively neglected in philosophy, under the assumption that they can be unproblematically treated as existential quantifiers. However, an important tradition in linguistic semantics, drawing from Stoic logic, draws out patterns which suggest that indefinites are not well treated simply as existential quantifiers.
There are two broad classes of response to puzzles like this, e-type and dynamic. These approaches raise deep foundational questions. Inter alia, both require revisionary notions of sentential content and non-classical treatments of the connectives. The proper treatment of (in)definites is thus of crucial importance to philosophical questions about the nature of content, the meaning of (in)definites, and the logic of natural language.
In this paper I develop a new approach to (in)definites. On my theory, contents are static, and indefinites have the truth-conditions of existential quantifiers. But they also have a secondary role: they have a witness presupposition which requires that, if the indefinite's truth is witnessed by any individual, then some such individual is assigned to their variable. This means that indefinites license subsequent anaphora to their witnesses. Crucially, the connectives in this system are classical. This shows that we can account for the behavior of (in)definites with resources that are much more conservative than those deployed by e-type or dynamic theories---and in particular, with a classical notion of content and connectives.
Eavesdropping judgments (judgments about truth, retraction, and consistency across contexts) about epistemic modals have been used in recent years to argue for a radical thesis: that truth is assessment-relative. We argue that judgments for 'I think that p' pattern in strikingly similar ways to judgments for 'Might p' and 'Probably p'. We argue for this by replicating three major experiments involving the latter and adding a condition with the form 'I think that p', showing that subjects respond in the same way to 'thinks' as to modals. This poses a serious challenge to relativist treatments of the modal judgments, since a relativist treatment of the corresponding 'thinks' judgments is totally implausible, so if a unified account of the phenomena is to be found, it cannot be a relativist one. We briefly sketch how a unified account might look.
This paper is about guessing: how people respond to a question when they aren't certain of the answer. Guesses show surprising and systematic patterns that the most obvious theories don't explain. We offer a theory that does explain them: we propose that people aim to optimize a tradeoff between accuracy and informativity in forming their guess. After spelling out our theory, we use it to argue that guessing plays a central role in our cognitive lives. In particular, our account of guessing yields new theories of (1) belief, (2) assertion, and (3) the conjunction fallacy: the psychological finding that people some-times rate conjunctions as more probable than their conjuncts. More generally, we suggest that guessing helps explain how boundedly rational agents like us navigate a complex, uncertain world.
The Identity principle says that conditionals with the form 'If p, then p' are logical truths.
Identity is overwhelmingly plausible, and has rarely been explicitly challenged.
But a wide range of conditionals nonetheless invalidate it.
I explain the problem, and argue that the culprit is the principle known as Import-Export, which we must thus reject.
I then explore how we can reject Import-Export in a way that still makes sense of the intuitions that support it, arguing that the differences between indicative and subjunctive conditionals play a key role in solving this puzzle.
My 2017 MIT dissertation, Coordination in Conversation, comprised four chapters which have now been published as
'Bounded modality', 'What 'must' adds', 'Modality and expressibility', and 'How to do things with modals'.