Accommodating Presuppositions

In this paper I elaborate on previous criticisms of the influential Stalnakerian account of presuppositions, pointing out that the well-known practice of informative presupposition puts heavy strain on Stalnaker’s pragmatic characterization of the phenomenon of presupposition, in particular of the triggering of presuppositions. Stalnaker has replied to previous criticisms by relying on the well-taken point that we should take into account the time at which presupposition-requirements are to be computed. In defense of a different, ‘semantic’ (in a sense) account of the phenomenon of presupposition, I argue that that point does not suffice to rescue the Stalnakerian proposal, and I portray Lewisian ‘accommodation’ as one way in which speakers adjust themselves to one another in the course of conversation.


Introduction
In this paper I elaborate on previous criticisms of the influential Stalnakerian account of presuppositions, pointing out that the well-known practice of informative presupposition puts heavy strain on Stalnaker's pragmatic characterization of the phenomenon of presupposition, in particular of their triggering. Stalnaker has replied to previous criticisms by relying on the well-taken point that we should take into account the time at which presuppositionrequirements are to be computed. In defense of a different, 'semantic' (in a sense to be explained in the first section) account of the phenomenon of presupposition, I argue in the second section that that point does not suffice to rescue the Stalnakerian proposal. In the final section I portray Lewisian 'accommodation' as one way in which speakers adjust themselves to one another in the course of conversation.

The Stalnakerian Picture
In a series of papers, Stalnaker (1973Stalnaker ( , 1974Stalnaker ( , 2002 has provided an influential account of the phenomenon of presupposition. The account has been slightly modified along the way; in this initial section I will just present what I take to be the core aspects relevant for my discussion. 1 Stalnaker's proposal is in the spirit of Grice's account of phenomena such as conversational implicature in particular and meaning in general: it purports to explain those phenomena as a specific form of rational behavior involving communicative intentions, avoiding irreducibly social notions such as conventions or (socially construed) norms. Stalnaker bases his analysis on a notion of speaker presupposition, which he then reluctantly (for reasons to be indicated presently) uses to provide a notion of sentence presupposition. The notion of speaker presupposition is explained in terms of common beliefs about what is accepted by the conversational partners; and common belief follows the pattern of Schiffer's and Lewis' proposals about it and about common knowledge: p is common M. García-Carpintero (&) LOGOS-Departament de Lògica, Història i Filosofia de la Ciència, Universitat de Barcelona, Barcelona, Spain e-mail: m.garciacarpintero@ub.edu belief in a given group G just in case (almost) everybody in G believes p, believes that (almost) everybody in G believes p, and so on. Acceptance is in its turn defined by Stalnaker (2002, 716) as a category of mental states ''which includes belief, but also some attitudes (presumption, assumption, acceptance for the purposes of an argument or an inquiry) that contrast with belief and with each other. To accept a proposition is to treat it as true for some reason.'' The need to invoke acceptance in the definition derives from many cases in which, intuitively, p is presupposed while not commonly believed. Thus, consider Donnellan's example: the secret conspirator asks the usurper's minions, ''Is the king in his countinghouse?'' Here the speaker does not believe that the intended referent is king, nor perhaps that there is a king, and hence does not believe that these propositions are commonly believed in the context, but nonetheless it is presupposed that the referent is king and that there is exactly one king. 2 Acceptance, however, cannot be invoked all the way down; the account is given in terms of common belief about what is commonly accepted, because only the more specific category of belief has the required explanatory links with behavior. This is thus the account. We first define a proposition p to be in the common ground in a group G-CG G (p)-and then we define speaker presupposition: (CG G ) CG G (p) if and only if it is common belief in G that everybody accepts p (SpP) Speaker S presupposes p (relative to G) if and only if S believes that CG G (p) Stalnaker (1973Stalnaker ( , 451, 1974 then defines a notion of sentence presupposition in terms of this: (SnP) Sentence S presupposes p if and only if the use of S would for some reason be inappropriate unless the speaker presupposed p Stalnaker (1978) complements this analysis of presuppositions with an analysis of assertion also deservedly influential, on which an assertion is a proposal to update the common ground, which, if accepted, is ''added'' to it (i.e., it then becomes common belief that every participant accepts it); and he combines the two accounts to suggest intuitively plausible explanations of some aspects of the projecting behavior we presented in the previous section. This (together with the related independent work of Lauri Karttunen) was the origin of the new important tradition of Dynamic Semantics (DS), developed for instance in Heim (1983), Beaver (2001) or von Fintel (2004. Unlike the traditional account of presuppositions as conditions on the truth and falsity of statements, DS can explain the selective projection behavior characteristic of presuppositions, and it can distinguish them from conventional implicatures. Geurts (1999, 17), however, is right in pointing out the important conceptual differences between the DS tradition and Stalnaker's viewpoint, which in fact go to the heart of the main issues I want to discuss here. Renouncing Stalnaker's Gricean reductive aims, in this tradition presuppositions are taken to be, both with respect to their triggering and projecting behavior, a constitutive feature of the semantics of natural language expressions. 3 Let us be a bit more clear and explicit about the differences between Stalnaker's ''pragmatic'' view and the ''semantic'' one I favor. As Stalnaker (1974, 61) notes, there are two contrasting ways of understanding the semantic/pragmatics divide. In the truth-conditional account, semantics deals with the truth-conditions of sentences, and the truth-conditional import of expressions. It is in this sense that presuppositions understood as conditions for the truth and falsity of sentences are said to be a semantic phenomenon. An important strand of Stalnaker's early defense of a pragmatic account, as he notes, is to oppose such a ''semantic'' conception; for well-known reasons (so-called ''non-catastrophic presupposition failure, 4 projection behavior''), I think that this opposition was well taken. However, as I have argued in detail elsewhere, 5 the truth-conditional way of tracing the semantic/pragmatic divide is not theoretically useful, because it displaces from the purview of semantics facts that should be studied together with those it keeps there: among others, semantically driven context-dependence, semantics for conventional indicators of speech acts such as the interrogative and imperative mood, and, indeed, (if the view to be promoted infra is correct) certain presuppositional facts.
On a different constitutive understanding of the divide, linguistics in general purports to theoretically characterize the constitutive facts about natural languages (in an indirect way of putting this, the linguistic competence of speakers), and semantics is the part thereof dealing with meaning facts constitutive of natural languages. This is, I think, the 2 I believe this is an intuitively correct characterization of what is presupposed in this case, which I take to be a referential use of the description; in general, as I argue in García-Carpintero (2000), all cases of reference involve ''identification'' presuppositions. 3 Geurts (1999, 14) distances himself from dynamic semantics on account of its betrayal of Stalnaker's truly pragmatic stance, and, like Stalnaker, he helps himself to a notion of expression-presupposition, defined in normative terms on the basis of the pragmatic notion of speaker presupposition. Unlike Stalnaker, Geurts also appeals to unexplained normative notions in characterizing speaker presupposition: ''a speaker who presupposes something incurs a commitment … regardless whether he really believes what he presupposes'' (ibid., 11). Geurts never explains where those requirements and commitments come from; they are prima facie at odds with the Stalnakerian stance he vows to adopt. 4 Cf. Yablo (2006). 5 Cf. García-Carpintero (2006). conception of the divide that Grice (1975) had in mind when he tried to account for the apparent asymmetric, nontruth-conditional behavior of conjunction or referential uses of descriptions as generalized conversational implicatures, i.e., as ''pragmatic'' features. Although his views here are complex, 6 I take this also to be Grice's (1981) own view on the presuppositional phenomena we are discussing here. After noting the two different interpretations of the divide, Stalnaker (1974, 61) points out that he is mainly arguing for a pragmatic account of presuppositions only on the first understanding, but notes also that his arguments have repercussions for the other: while he is open to the possibility that in some cases ''one may just have to write presupposition constraints into the dictionary entry for a particular word'' (ibid.), he conjectures ''that one can explain many presupposition constraints in terms of general conversational rules without building anything about presuppositions into the meanings of particular words or constructions'' (ibid.).
In fact, although as we have seen Stalnaker introduced a notion of sentence presupposition (SnP) in his early writings, and still assumes it in recent writing, he repeatedly expresses qualms about it, because of the unexplained appeal to the normative notion of inappropriateness, and because it suggests the existence of a ''mysterious relation X'' between sentences and propositions worthy of analysis, while ''we don't need the mysterious relation X to describe the phenomena, and it does not make any contribution to explaining them'' (2002,(712)(713). 7 We may say that Gricean generalized conversational implicature accounts of referential uses of descriptions or manifest non-truth-conditional asymmetries in conjunctions are not simply reductionist, but in fact eliminativist vis-à-vis semantic accounts of those phenomena, on the second understanding of the divide: although it is acknowledged that definite descriptions and conjunctions are in fact commonly used in those ways, it is claimed that a semantic theory should not encompass them. This is the way I understand in this paper the label '(Gricean) eliminativist view' of the phenomenon here studied, presupposition, and apply it to writers such as Boër and Lycan (1976), Levinson (1983) and, indeed, Grice (1981). The proposal is not to deny the phenomenon altogether, but only the need for a semantic account for it. Presuppositions do exist, but they can be accounted for without including them in our theoretical constitutive characterization of natural languages. The Stalnakerian view of presuppositions, in contrast with the DS view, is ultimately eliminativist in this sense. 8 This stance was present from the beginning, but the emphasis is stronger in more recent work: [O]ne might define a notion of sentence presupposition in terms of speaker presupposition, but […] the attempt to do so would be a distraction, and would not yield any theoretically useful notion (2010, 150).
In the next section I will critically examine these contentions. While I will essentially agree with Stalnaker that presupposition is a pragmatic, not semantic phenomenon in the truth-conditional sense, ultimately having to do with the propositional attitudes of speakers, I will provide reasons to question his Gricean eliminativist stance, and hence to reject that it is a pragmatic phenomenon also on the constitutive account.

The Problem of Accommodating Accommodation
As Stalnaker (1973Stalnaker ( , 449, 1974 noted in his early writings, it is common for speakers to communicate a piece of information by uttering a sentence that presupposes it. These are real life examples from Abbott (2008, 531, cf. sources there):

The leaders of the militant homophile movement in
America generally have been young people. It was they who fought back during a violent police raid on a Greenwich Village bar in 1969, an incident from which many gays date the birth of the modern crusade for homosexual rights. 2. If you're going into the bedroom, would you mind bringing back the big bag of potato chips that I left on the bed?
Speakers who utter sentences (1) and (2) do not typically assume their presuppositions-that some people fought back during a violent police raid on a Greenwich Village bar in 1969, and that there is exactly one big bag of potato chips that the speaker left on the bed, respectively-to be in the common ground. To utter sentences with those presuppositions is just an expedient resource for them to inform their audiences of such contents, plus the assertion, woven together in a terse package. Most writers, however (Stalnaker among them), would like to count the relevant contents as nonetheless somehow presupposed. 9 The examples above help us to appreciate the ordinariness of the phenomenon, but usually simpler cases are discussed. 10 We assume that the speaker utters (3) in the knowledge that his audience knows nothing about his family: 3. I cannot come to the meeting-I have to pick up my sister at the airport.
As in the previous case, speakers utter sentences like (3) as a convenient resource for them to inform their audiences of such contents, plus the assertion. Stalnaker, however, like most other writers, would like to count the relevant contents as nonetheless somehow presupposed. These are cases where speakers exploit what Lewis (1979) called the ''Rule of Accommodation for Presuppositions'', which he characterized thus: (RA) If at time t something is said that requires presupposition p to be acceptable, and if p is not presupposed just before t, then-ceteris paribus and within certain limits-presupposition p comes into existence at t.
Cases in which the sort of hey, wait a minute complaint that von Fintel (2004, 271) (citing Shannon 1976) proposes as a test to distinguish presupposition and assertion is made explain the need for the hedge: the hearer is not always prepared to accommodate. Thus, imagine that the speaker had made the utterance with 'my lover' replacing 'my sister': the audience might well have then objected, ''hey, wait a minute, I did not know you had a lover''. Now, the initial problem for Stalnaker's account that cases of informative presupposition pose is as follows: (1) as he (1973,449,1974, acknowledges, a presupposition is present; 11 however, (at first sight at least) (2) the speaker does not presuppose it, on Stalnaker's characterization, because he does not believe that his audience accepts it; while (3) the fact that cases like these are commonplace suggests that there is nothing inappropriate in their use, and certainly nothing feels inappropriate in them. 12 Although he has been aware of the issue all along, only in recent work has Stalnaker confronted it squarely, arguing that in fact these cases are not at odds with his account, because only at first sight is (2) correct: when the proper time at which the presupposition is to be accepted is considered, it turns out that the speaker is presupposing the relevant content. I will now critically discuss the adequacy of his arguments, which writers sympathetic to Stalnaker's pragmatic account such as Simons (2003, 267-268) and Schlenker (2012) endorse. 13,14 Stalnaker (2002, 708-709) points out that utterances themselves are manifest events, which become part of the common ground. Given that speakers take advantage of this, speakers' presuppositions should only be satisfied at a ''(perhaps somewhat idealized) point after the utterance event has taken place, but before it has been accepted or rejected''. Stalnaker (1998, 101) motivates this with a convincing example: The point of a speech act […] is to change the context, and since the way the speech act is supposed to 9 On the view that I will suggest below, an informative presupposition is a pragmatically created one: the speaker uses a device that conventionally presupposes something in order to get the speaker to presuppose it. The main reason for acknowledging the presence of a presupposition here, to which Stalnaker is sensitive, lies in the aim to provide a systematic compositional account of their semantics. 10 Stalnaker (1974, 52, n. 2) attributes the following example to Jerry Sadock. 11 Kadmon (2001, 219-221) describes these instead as cases of presupposition ''disappearance'', on the basis of her characterization of presuppositions as propositions ''intuitively felt to be taken for Footnote 11 continued granted''. But I think this is a bad choice, based on an inadequate, manifestly overgenerating characterization. 12 Abbott (2008) and Gauker (2008, 185) make critical points related to the ones presented below. They, however, contend that the appeal to accommodation to deal with informative presupposition renders any common knowledge account of presuppositions vacuous. On the argumentative line I will sketch, the phenomenon poses problems to pragmatic views such as Stalnaker's, but accounts that assume semantic triggering such as DS can surmount them; I fail to see how claims of vacuity can be substantiated against views of that shape. 13 Simons (2003, 267-269), who shares Stalnaker's eliminativist leanings, shows that more complex sentences may pose difficulties for Stalnaker's ''idealized time'' strategy that is described below. In more recent work ( 2010.), in which she deepens her eliminativist viewpoint, she doubts that the strategy might suffice to account for informative presuppositions, on the basis of considerations related to the ones developed below. 14 In his insightful discussion of accommodation (to which I am much indebted), von Fintel (2008) makes heavy use of Stalnaker's point about the proper time at which presuppositions should be satisfied by the common ground. However, his view of presuppositions differs from Stalnaker's precisely on the matter we are discussing: he accepts semantically triggered presuppositions (cf. p. 138). There is no problem at all in accepting that speakers do make the relevant presuppositions, invoking for it Stalnaker's claim about the time when they should be accepted, if in the cases we are discussing they are semantically triggered-but only under that condition. What is problematic, and I will be questioning, is Stalnaker's claim that the same applies even if there is no linguistic trigger in the sentences. My account of accommodation in the next section is, I think, close to von Fintel's. change the context depends on its content, interpretation must be done in the prior context -the context as it is before the assertion is accepted, and its content added to what is presupposed. But the prior context cannot be the context as it was before the speaker began to speak. Suppose Phoebe says 'I saw an interesting movie last night'. To determine the content of her remark, one needs to know who is speaking, and so Phoebe, if she is speaking appropriately, must be presuming that the information that she is speaking is available to her audience -that is shared information. But she need not presume that this information was available before she began to speak. The prior context that is relevant to the interpretation of a speech act is the context as it is changed by the fact that the speech act was made, but prior to the acceptance or rejection of the speech act. 15 Stalnaker thus rejects (2) in the characterization of the problem above: the attitudes constitutive of speaker presupposition on his account were in fact there-not relative to the time before the speaker made his utterance, which is not after all when they should be, but to the ''somewhat idealized time'' at which they are required to be in place. Now, although Stalnaker's contention about the time when the speakers' attitudes he takes to be constitutive of presuppositions should be (ideally) present is undoubtedly correct-as the Phoebe example clearly shows-this, by itself, does not suffice to account for informative presuppositions in an eliminativist setting; for simply pointing this out does not suffice to explain how they have been triggered by then, which is what is at stake.
For our present purposes, the two times in idealized interpretation-one after the utterance ends, when the presuppositions are checked, before the second one at which acceptance or rejection of the assertion is decidedexist insofar as the presuppositions do: presuppositions are just those contents considered at the first moment in ideal interpretation. 16 What is at stake in this debate, however, is whether there are cases in which they are semantically triggered. Not in all cases are they; there clearly are pragmatically triggered presuppositions, for which the two ideal moments nonetheless also exist (as when somebody tells me, ''you surely celebrated all night'', assuming that my club won the cup). What is here in question is whether their presence can be accounted for pragmatically in all cases, including those that at first sight require a semantic trigger, such as the 'my sister' case. Stalnaker assumes the presence of the two moments in all cases, including those, without a semantic trigger (a ''mysterious relation X''); in so doing he begs the question whether or not, in the cases under dispute, the existence of the two separate moments in idealized interpretation can be duly justified.
Remember that what is at stake is whether informative presuppositions are compatible with the Gricean eliminativist stance that Stalnaker ultimately professes. Even if he declares himself open to the existence of conventional triggers, as we saw he (2002, presses for the eliminativist view: ''Suppose we assume that the semantics tells us exactly this about the sentence 'I have to pick up my sister at the airport': it is true if and only if the speaker has a sister whom he or she has to pick up at the airport, and false otherwise. So we are supposing that the semantics tells us nothing either about relation X, or about what speakers must take to be common ground. Are there facts about the use of the sentence that cannot be explained by this semantic hypothesis, together with general conversational rules?'' This is a rhetorical question. But the facts of informative presupposition suggest that the answer is ''yes'', disappointing the rhetorically conveyed expectations. The Phoebe example does support Stalnaker's claim about the time when speakers should believe that their presuppositions are accepted; but, in the context of our dialectics, the problem for him is that the presupposition in that example (that x is the speaker who uttered ''I'') seems as much conventionally (semantically) triggered, as the one in (3) we are discussing. It rings true that, at a ''somewhat idealized time'' after the utterance, before acceptance or rejection of the main assertoric claim-that x saw an interesting movie last night-, the presupposition that x is the speaker who uttered 'I' must be accepted. But this appears to be induced by the semantics of the sentence, and this is certainly what partisans of Stalnaker's rival account of presuppositions would want to say. 17 In the context of the present dialectics, Stalnaker should rather have considered an utterance of, say, 'there is exactly one agent of this very utterance, and s/he saw an interesting movie last night' (analogous to the semantically given truth-conditions he considers for (3) in the quotation two paragraphs back, ''Suppose we assume that the semantics tells us exactly this …''). Would an ideal interpretation of that utterance also require two different interpretative ''moments'', an earlier one at which a speaker presupposition identifying an individual as the 15 The point was in fact made earlier by Stalnaker (1978, 86): ''the context on which assertion has its essential effect is not defined by what is presupposed before the speaker begins to speak, but will include any information which the speaker assumes his audience can infer from the performance of the speech act''. 16 Of course, information in addition to presuppositions is also added at the first stage; consider, say, that Phoebe is speaking, which is not a presupposition.
17 García-Carpintero (2000) promotes such a view about the semantics of indexicals, demonstratives and proper names. speaker is checked and a later one at which acceptance or rejection of the claim that that individual saw an interesting movie is decided? I do not see why: in this case, those two propositions are part of a single assertoric content, which is to be considered at the second time. By parity of reasoning, the same applies to the case in dispute of (3) on Stalnaker's assumptions about what the semantics tells us. Simply taking for granted that the ideal moment in question exists even on eliminativist assumptions begs the question at issue: for it simply assumes, without independent justification, that the presuppositional requirement is somehow triggered even without conventional indicators.
In sum: Stalnaker is right that when an utterance triggers (semantically or pragmatically) a presupposition, proper evaluation of it involves two moments, an earlier ''presupposition moment'' PM at which the presuppositions are checked, a later ''assertion moment'' AM at which the assertoric content is appraised. As an utterance of (3) triggers the presupposition that the speaker has a sister, this is to be in place at PM for it to be correct. Nevertheless, pointing this out does not address the issue at stake, which is how the presupposition was triggered, whether lexically (semantically) or pragmatically. In fact, if we consider instead of (3) a sentence straightforwardly expressing what Stalnaker takes to be the semantically conveyed content-that the speaker has a sister whom he or she has to pick up at the airport-there does not seem to be anything in need to be checked at PM, which suggest that the triggering in (3) was after all semantic, and, given that this is compatible with acknowledging the need to separate PM and AM, reinforces the need for an argument that the triggering is after all pragmatic. Hence, just pointing out the need to distinguish PM from AM does not address the question at stake, and in assuming that it does Stalnaker begs it. Now, in reply to Abbott (2008), who also mentions in response to Stalnaker's rhetorical question quoted above the case of informative presuppositions as facts that cannot be explained by his semantic hypothesis, Stalnaker (2008, 542) replies: […] the general account of assertion does explain this difference. The assumption is that an assertion is something like a proposal to add the information that is the content of the assertion to the common ground, and a rejection of the proposal is a normal move in the conversational game. Accommodated information is communicated indirectly, so that there is no provision for straightforwardly rejecting it. (One has to say something like ''Hey, wait a minute'' -one of the tests that Kai von Fintel has used to identify presupposition.) That is why accommodated information survives rejection, and it is why it is inappropriate to communicate information that is either controversial or noteworthy by presupposing it.
However, if 'I have to pick up my sister at the airport' and 'I have a sister whom I have to pick up at the airport' are semantically on a par, how come the information that the speaker has a sister is only ''communicated indirectly'' in the former case but not in the latter, so that it survives rejection in the former case but not in the latter in the way Stalnaker explains? In assuming that it is merely indirectly communicated, without telling us why that is so given his views, Stalnaker is once more just begging the question at stake. What fact grounds the distinction he is appealing to? Within his framework, there appears to be no resources to explain it. The truth-conditions are the same; the prior context is the same; the background assumptions about what speakers can normally be expected to know are the same.

Accommodation in a Semantic Account of Presupposing
As I have already mentioned, although accounts of presupposition such as Heim's (1983), Beaver's (2001) and even perhaps Geurts's (1999) are in the spirit of Stalnaker's, they straightforwardly abandon his eliminativist leanings. Presupposing is a pragmatic notion in the truthconditional sense, but not in the constitutive one, involving attitudes of speakers. Presuppositions can be triggered in different ways, even when standardly associated with sentence-types-allowing for conventional triggering in such cases, but not requiring it. These accounts explicitly assume that there are conventional indicators of presuppositions (such as the cleft construction, or definite descriptions), which a semantic theory should properly countenance, perhaps in its lexical semantics component.
To give an account of what happens in the case of informative presuppositions on the basis of such assumptions is relatively straightforward. Conventionally (and semantically, on the constitutive view of the semantics/ pragmatics divide), whoever utters 'Who the heck would want to see that film?' is asking a question, but the speaker is in fact asserting something. Conventionally, and semantically, an utterance of 'Paul is a good friend' is an assertion that Paul is a good friend; in some contexts, it might be perfectly clear that the speaker is not making such an assertion, but in fact one with a contrary content. Conventionally, and semantically, 'Thanks for not browsing our magazines' is an expression of gratitude, but when we find an utterance of it in the train station kiosk, we know that its author was doing no such thing, but in fact making a request. Conventionally, and semantically, the sentence that George Eliot writes at the beginning of Middlemarch ''Miss Brooke had that kind of beauty which seems to be thrown into relief by poor dress'' presupposes the existence of a specific 'Miss Brooke' naming practice, on which she relies for the purpose of identifying a person about whose beauty she makes a claim; but Eliot is not making either the presupposing or the asserting, she is just putting her audience in a position to imagine something. 18 Something similar happens in cases of informative presuppositions. Speakers who utter ''I cannot come to the meeting-I have to pick up my sister at the airport'' or ''I am sorry I am late-my car broke down'' use sentences that conventionally and semantically presuppose that they have a sister and a car, respectively, and they themselves as speakers might be presupposing it: for instance (if this is what presupposing ultimately is, which I have not discussed here), they might be assuming that they are performing an act-ancillary to their main speech act-which is correct if, and only if, it is mutually known to all involved that he has a sister, or a car. But in the cases of informative presuppositions we have been discussing, to the extent that they are presupposing in this sense, this is essentially because they are relying on their audiences' awareness of the presuppositional requirement conventionally attached to the form of words they are using, and their awareness that in their context the requirement prima facie is not met, the speaker knows that it is not, knows that the audience knows that it is not, and so on and so forth, in order to indirectly provide them those pieces of non-controversial background information in a conveniently brief and non-verbose way. 19 We can rely on Stalnaker's distinction between PM and AM to argue that the presupposition was after all present, to be satisfied only at PM, assumed to occur at the right time; but it is our relying on semantic triggering that crucially allows us to do that without begging any issue. This is thus my diagnosis of the case of informative presuppositions, vis-à-vis the triad I used in the previous section to describe Stalnaker's difficulty: (1) semantically, a presupposition is present; (2) the speaker is presupposing, in the way Stalnaker suggests, but (unlike him) by assuming a semantic triggering; (3) there is nothing inappropriate in it, exactly in the way that there is nothing such in all those analogous examples of indirection we have just mentioned. 20 As Karttunen (1974, 412) puts it, ''This is one way in which we communicate indirectly, convey matters without discussing them.' ' Stalnaker (1974, 51-52;cf. also 1973, 451) also accounts for these cases essentially in this way: ''In such a case, a speaker tells his auditor something in part by pretending that his auditor already knows it''. What I have argued is that there are serious objections to his claim that such a ''pretending'' can be accounted for if there is no ''mysterious relation X'', i.e., that Stalnaker cannot stick to his quoted diagnosis consistently with his reductionist leanings.
Simons (2010) has recently provided an eliminativist account alternative to Stalnaker's. On her account, the basic concept is utterance presupposition; both speaker presupposition, and an etiolated notion of sentence presupposition along the lines of Stalnaker's are defined in terms of it. She defines utterance U presupposes p thus: (1) It is not part of the speaker's primary communicative intention to convey p, and (2) the interpreter of U must take the speaker of U to accept p in order to make sense of U. Now, prima facie informative presuppositions-especially those in which the intuitively primary point of the speaker is to convey the presupposition, as in the notorious exchange: 'The new boss is attractive -yes, his wife thinks so too'-21 constitute a counterexample to this proposal. Simons deals with this by explaining that she does not mean 'primary' in any intuitive sense, but in a technical one: on her view, in such cases ''the speaker is being intentionally indirect, and is exploiting the presuppositional requirements of the utterance […] when a speaker produces an utterance with the specific intention of communicating what is presupposed, this intention must be viewed as a secondary intention. This is why the definition above refers to the speaker's primary communicative intention'' (op. cit., 20-1). Now, on the sort of semantic proposal I have been making, Simons' primary intention is not at all difficult to cash out: it is just the intention of conveying the conventional/semantic content of the utterance, and the previous Gricean account elaborates in which ways the speaker is 18 I take it that these are all examples of indirect speech acts, which is how I propose to understand informative presuppositions (assuming presupposition, like reference, to be an ancillary speech act); cf. García-Carpintero (2013). Some of them are described in the literature as non-literal uses-a category I would rather reserve for metaphors and loose talk. Whatever the best classification of the preceding examples is, I would suggest to include in it informative presuppositions. 19 The proposal thus provides an elaboration or explanation of Lewis' RA, the Rule of Accommodation for presuppositions. The way ''presupposition p comes into existence at t'' is by its being added to the common ground at PM. 20 This is, I take it, the picture presented by von Fintel (2008, 151). 21 Alan Ryan's review of John Stuart Mill: Victorian Firebrand (''The Passionate Hero, Then and Now'', New York Review of Books, 2011, 19, 60) contains the following quotation from the book with a nice real life example; it refers to Mill's first encounter with Harriet Taylor, who would become his very special friend for 20 years until the death of her husband, and then his wife: ''In many ways, it was not a surprising match. Harriet Taylor was intelligent, pretty, vivacious, progressive, open-minded and poetic. But his admiration was shared by others-not least by her two children, and her husband''. I assume that this is the first indication in the book that Harriet was married and had two children.
being indirect in secondarily conveying the informative presupposition. 22 In other words, such a semantically based account has the resources for non-circularly explaining the crucial notion that Simons appeals to, that of exploiting the presuppositional requirements of the utterance. But she does not explain how the same result can be obtained given her eliminativist stance. The problem is how to establish, without the semantic assumption, that conveying that the new boss has a wife is not part of the ''speaker's primary communicative intention'', in her technical sense. It is not, she says, because it is part of the ''presupposition requirements'' of the utterance; but where do such requirements come from, if they are supposed to consist merely of psychological attitudes of the speakers and their interpreters? This looks like one more eliminativist proposal that begs the main question at stake.