Abstract
Strawson (in Mind 59:320–344, 1950; Theoria 30(2):96–118, 1964) argued that definite NPs trigger presuppositions as an aspect of their conventional meanings, and this semantic conception of presupposition triggers is incorporated into the binding theory of presuppositions (van der Sandt, in J Semant 9:333–377, 1992 and Geurts, in Presupposition and pronouns, 1999). The phenomenon of presupposition denials, however, presents a problem for the semantic conception of presupposition triggers, for in such denials the alleged semantic presuppositions seem to be “cancelled” by a negation operator. Geurts (in Language 74(2):274–307, 1998; Presupposition and pronouns, 1999) attempts to solve this problem by utilizing the binding theory’s allowance for local accommodation. Geurts’ proposal, however, is inadequate, primarily because Geurts formulates the binding theory within discourse representation theory (DRT), which makes no allowance for the illocutionary force of denial. Asher and Lascarides’ (in J Semant 15:239–299, 1998a; J Semant 15:83–113, 1998b; The logics of conversation, 2003) segmented discourse representation theory (SDRT), however, has the resources to account for the illocutionary force of denial. Utilizing a version of SDRT I demonstrate that the “cancelling” of a presupposition by echoic negation in presupposition denials crucially depends upon the denial of information that is pragmatically communicated by accommodation.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Here I will be concerned only presuppositions triggered by definite NPs, including proper names. The extent to which presupposition triggers form a uniform class, and thus the extent to which presupposition is a unified phenomenon, is an open question that will not be addressed here. For recent discussion see Beaver and Geurts (2013), Roberts et al. (2009), Cummins et al. (2013), and Tonhauser et al. (2013).
Presupposition failure does not always give rise to intuitions of infelicity. Determining when presupposition failure gives rise to judgments of infelicity, as opposed to falsity, is a subtle matter which will not be addressed here. For some interesting proposals see Von Fintel (2004), Yablo (2006), Geurts (2007), and Schouebye (2009).
Several caveats are in order. First, I am not claiming that the binding theory accounts for all phenomena that have been classified as presuppositional; there are several outstanding problems for the binding theory, in addition to the problem of presupposition denials. Second, I am not claiming the binding theory is the only viable theory of presupposition. [Beaver (2001) and Schlenker (2011) present phenomena that are problematic for the binding theory, and they present compelling entailment-based accounts.] I am claiming merely that the binding theory is a compelling approach to the projection problem, especially for presuppositions triggered by definite NPs.
The qualifier “in a way” is appropriate because the contradiction is not classical contradiction: the truth of \((\hbox {A}_{3})\) does not imply that \((\hbox {A}_{2})\) is false, but rather that \((\hbox {A}_{2})\) is infelicitous. Burton-Roberts (1989a, b) argues that the phenomenon of presupposition denial actually supports the semantic conception of presupposition triggers because presupposition denials are instances of Horn’s (1985) metalinguistic negation, and this requires that \((\hbox {A}_{2})\) and \((\hbox {A}_{3})\) be contradictory. Carston (1998, 2002), however, demonstrates that the argument relies on an overly simplistic view of metalinguistic negation, and she advances a purely pragmatic account of presupposition denials.
A tempting response on behalf of the semantic view is to posit a lexical ambiguity in negation: one sense of ‘not’ (“choice negation”) would allow presuppositions to project through it, while another sense (“exclusion negation”) would block the projection of presuppositions. There is, however, no independent evidence in support of this ambiguity hypothesis.
Presupposition denials are problematic not only for the biding theory, but for the semantic approach to presupposition triggering generally. Beaver notes that presupposition denials are problematic for his entailment-based account, and he briefly canvasses several potential solutions. He rejects them all, however, because each of them “would not make sense of what it is to deny, would not explain or account for the distribution of such denials in discourse, and further would not account for the observation that what can be denied is not limited to presuppositions and assertions” (2001, p. 273). The solution to the problem of presupposition denials advanced here does not suffer from these shortcomings.
The account of presupposition denial developed here is inspired by, yet superior to, the accounts developed by Van der Sandt (1991), Van der Sandt and Meier (2003), and especially Asher and Lascarides (2003). Limitations of space prevent me from considering these proposals in detail, but as I proceed I will note, in footnotes, the ways in which my account improves upon these preceding proposals.
Geurts (1999, p. 42) defines employment as follows:
Emp(\({\upphi }\)) is the smallest set of reference markers for which the following hold:
-
a.
If \(\hbox {Pu}_{1}\ldots \hbox {u}_{\mathrm{n}} \in \hbox {Con}({\upphi })\), then \(\{\hbox {u}_{1}\ldots \hbox {u}_{\mathrm{n}}\} \subseteq \) Emp(\({\upphi }\))
-
b.
If u = v \(\in \) Con(\({\upphi }\)), then {u ,v} \(\subseteq \) Emp(\({\upphi }\))
Properness is then defined in terms of \(\le \), Emp( ), and Acc( ): DRS \({\upphi }\) is proper iff \(\forall {\uppsi } \ge {\upphi }\): Emp\(({\uppsi }) \subseteq \textit{Acc}({\uppsi }\)).
-
a.
Geurts’ version of clause (a) is more complex: \(\hbox {M}\models _{f} {\upphi }\) iff \(\exists g \supseteq f{:}\; \hbox {dom}(g) = \hbox {Acc}({\upphi })\) and \(\forall {\uppsi } \in \hbox {Con}({\upphi }){:}\; \hbox {M}\models _{g}{\uppsi }\). The added complexity has the effect of enforcing the stipulation that “no f can be a verifying embedding for \({\upphi }\) unless \(\hbox {dom}(f) = \hbox {Acc}({\upphi })\)” (1999, p. 42, note 2). As this additional complexity does not serve my purposes, I have opted for a simpler version of clause (a).
Geurts (1999, p. 57) does not restrict the first principle to only presuppositions triggered by definite NPs, but here I am concerned only with this class of presupposition trigger.
If the empty background is taken to reflect the fact that we eavesdropping theoreticians are ignorant of the mutually shared assumptions of the participants, this assumption is plausible. But if the assumption of an empty background is presented as a representation of the information that is mutually assumed by the discourse participants, it is implausible. One can always talk, in a shared language, about the weather. This distinction between the interpretation of we eavesdroppers and the interpretation of the discourse participants will be developed in Sect. 4.
Following Geurts (1999), I here rely on an intuitive construction of semantically encoded DRSs, and I suppress much irrelevant detail. Kamp and Reyle (1993) assume a standard generalized phrase structure grammar and present a much more detailed description of the procedure whereby DRSs are constructed from syntactic representations.
A linking of presupposed reference marker \(\tau \) to an accessible reference maker \(\rho \) is appropriate only if the conditions on \(\tau \) are at least compatible with the conditions on \(\rho \), though typically there is a significant degree of redundancy between the conditions. The existence of more than one appropriate resolution gives rise to ambiguities. The examples of binding considered here will involve unique and intuitively appropriate resolutions.
Later on I will demonstrate that if an utterance of \((\hbox {C}_{1})\) is taken as a correction, as opposed to a simple assertion, then there is no difficulty in discerning its illocutionary point, even if the presupposition is interpreted as being accommodated “locally,” inside the negation.
Entailment-based accounts can account for why \((\hbox {E}_{1})\) does not intuitively presuppose that Harry has a wife, but such accounts run into difficulty if the disjuncts are reversed: an utterance of ‘Either Harry’s wife is very tolerant, or he’s a bachelor’ also intuitively does not presuppose that Harry has a wife, but the latter intuition is not readily explained by entailment accounts. See Chemla and Schlenker (2012) for an empirically informed discussion of the symmetry issue.
What Geurts says here informally is captured in van der Sandt’s (1992) formal (and perhaps less perspicuous) definition of admissible resolution. Condition (iii) of the following definition renders the accommodation of the presupposition triggered by ‘Harry’s wife’ in (E) in the main DRS inadmissible:
Let \(\hbox {K}_{0}\) be the incoming DRS, \(\hbox {K}_{1}\) the merge of a DRS with \(\hbox {K}_{0}\) and \(\hbox {K}_{1}\)’ a possible resolution of \(\hbox {K}_{1}\). The resolution of \(\hbox {K}_{0}\) to \(\hbox {K}_{1}\)’ is subject to the following conditions in order to be admissible:
-
(i)
\(\hbox {K}_{1}\)’ is informative with respect \(\hbox {K}_{0}\), that is \(\hbox {K}_{0}\) does not entail \(\hbox {K}_{1}\)’.
-
(ii)
Resolving \(\hbox {K}_{0}\) to \(\hbox {K}_{1}\)’ maintains consistency.
-
(iii)
Resolving \(\hbox {K}_{0}\) to \(\hbox {K}_{1}\)’ does not give rise to a structure in which
-
(a)
some subordinate DRS \(\hbox {K}_{\mathrm{i}}\) is entailed by the DRSs which are superordinate to it
-
(b)
\(\lnot \hbox {K}_{\mathrm{i}}\) us entailed by the DRSs with are superordinate to it.
-
(a)
-
(i)
DRT also predicts that “intermediate accommodation” is possible, though the accuracy of this prediction is disputed by Beaver (2001). I will not examine the controversy over intermediate accommodation here.
Though denial is typically indicated by such discourse particles, it is not necessary to use such any such discourse particle to felicitously perform a denial.
Based in part on a mistaken belief that presupposition denials pose no special problem for the binding theory of presupposition, Geurts (1998) “repudiates the notion that denial is a phenomenon amenable to a unified treatment,” and he instead claims that the “underlying mechanisms” of proposition denials, implicature denials, form denials, or a presupposition denials “are quite diverse” (Geurts 1998, p. 277). Appreciation of the non-monotonic illocutionary effects of all four types of denial suggests that this eclectic stance toward denial is incorrect.
Van Leusen (2004), following Asher (1995), refers to sequences of utterances such as \((\hbox {A}_{2})\) followed by \((\hbox {A}_{3})\) in exchange (A) as corrections, and she argues that “for a discourse contribution to be felicitous as a correction, there must be a contextually supported alternative to the corrective claim such that the two are inconsistent given the context of interpretation” (2004, p. 2). Van Leusen argues that the felicity of both the denial (e.g. (\(\hbox {A}_{2})\)) and what I refer to as the follow-up (e.g. (\(\hbox {A}_{3})\)) require this contextually supported incompatibility.
One might suggest that this incorrect prediction could be avoided if the applicability of the pragmatic constraint of consistency were extended so that it applied not only to the process of accommodation, but could also be applied to block putative bindings of presuppositions. The idea would be that if the consistency constraint were extended in this way, the presupposition triggered by ‘Harry’s wife’ in \((\hbox {A}_{2})\) would be blocked from being bound to accessible and appropriate discourse referent w, and would instead be accommodated either locally, or globally. But extending the consistency constraint in this way will not solve the problem. Accommodating the presupposition locally, inside the negation, would also yields an inconsistent DRS; the local accommodation of the presupposition would yield a DRS that is true if and only if, speaking loosely, Harry has a wife who is rich, and Harry does not have a wife who is rich. Accommodating the presupposition globally would yield a consistent interpretation, but not an interpretation that is actually available: the global accommodation of the presupposition triggered by ‘Harry’s wife’ in \((\hbox {A}_{2})\) would yield a DRS that is true if and only if, speaking loosely, Harry has two wives, only one of whom is rich.
The semantics also requires that we introduce connectives ‘&’ and ‘\(\lnot \)’ for SDRS formulae. How this works is straightforward, so I will not belabor the point here.
Note that \(\hbox {K}_{\pi I}\) will either be atomic SDRS formula of the form \(R(\pi _{k,}\pi _{l})\), or a complex negated SDRS formula \(\lnot \varphi \) or a complex conjunctive SDRS formula (\( \varphi \, \& \, {\psi }\)). Thus label \(\pi _{{i}}\) immediately outscopes label \(\pi _{j}\) iff (i) \(\hbox {K}_{\pi I}\) is of the form \(R(\pi _{k,}\pi _{l})\) and either \(j=k\) or \(j=l\), or (ii) \(\hbox {K}_{\pi I}\) is of the form \(\lnot \varphi \) and \(\varphi \) contains some atomic formula \(R(\pi _{k,} \pi _{l})\) such that either \(j=k\) or \(j=l\), or (iii) \(\hbox {K}_{\pi I}\) is of the form (\( \varphi \, \& \, {\psi }\)) and either \(\varphi \) or \({\psi }\) contains some atomic formula \(R(\pi _{k,} \pi _{l})\) such that either \(j=k\) or \(j=l\).
It would be more accurate, though perhaps less helpful, to say that the box-diagrams are representations of SDRSs. More precisely defined, an SDRS is a triple \(\langle \hbox {A}, {\uppi }_{\mathrm{LAST},} \mathfrak {I}\rangle \), where A is a set of labels, \({\uppi }_{\mathrm{LAST}}\), is a label in A, and \(\mathfrak {I}\) is an assignment of labels in A to SDRS formulae. (Intuitively, \(\pi _{\textit{LAST}}\) is the label corresponding to the most recent utterance in the discourse.)
Because \((\hbox {BG}_{\mathrm{F}}^{0})\) contains no r-conditions, the semantics provided below entail that it is satisfied under all models, under all embedding functions. This is a somewhat odd consequence, but it is hardly counterintuitive.
Asher and Lascarides (2003) define some rhetorical relations (e.g. background) such that the attachment point for the under-specified \(\pi _{?}\) can be either the left or right argument. Thus their underspecified r-conditions have a disjunctive form: \(R_{?} (\pi _{?,} \pi _{1}) \vee R_{?} (\pi _{1,}\pi _{?})\). For the sake of simplicity I will assume that all rhetorical relations are defined in such a way that the “attachment point” is the left argument.
The semantics of elaboration( , ) also entail that the event of the dog barking must be included within the event of the noisy night. For sake of simplicity I am suppressing SDRT’s treatment of such temporal information.
My construal of SDRS update is more simplistic than the detailed account provided in Asher and Lascarides in yet another way: Rather than have the update operation yield a unique SDRS, the procedure defined by Asher and Lascarides (2003, p. 218) produces a set of SDRS, where the members of the set represent potential interpretations, or “live options.” These potential interpretations are then ranked in partial preference ordering, and the “winners” of the ranking are checked against our intuitive interpretations. Asher and Lascarides’ construal of update is more realistic than the simplistic idealization I am assuming here, but the “live options” construal introduces much complexity, and its advantages are independent of my central concerns.
An orthographic and technical note: Symbols such as ‘\((\hbox {BG}_{\mathrm{P}}^{0})\)’ belong to the metalanguage, and are used to refer to SDRSs, which are themselves representations formulated in the language of SDRT (what Asher and Lascarides call the language of information content). In contrast, ‘\({\uppi }_{\mathrm{BGP0}}\)’ is a symbol in the language of information content; it is a label of the SDRS displayed below. In an effort to keep things as simple as possible, I use labels of background SDRS, e.g. ‘\({\uppi }_{\mathrm{BGP0}}\)’, only when applying the semantic rules.
Asher and Lascarides (2003) provide a dynamic semantics for both DRT and SDRT. Simplifying somewhat, under such a dynamic semantics formulae are assigned as semantic values functions from sets of input assignment functions to sets of verifying assignment functions. In contrast, the DRT and SDRT semantics I provide is static; it simply provides necessary and sufficient conditions for a formulae’s being verified in a model by an assignment function. The static semantics provided here is more readily checked against intuitive judgements, and it meshes with Geurts’ (1999) static DRT semantics.
Setting aside some irrelevant details, Asher and Lascarides (2003, p. 353) propose that \(\hbox {M} \models _{f}\) correction( \(\pi _{{i}}, \pi _{j}\) ) iff \( \hbox {M} \models _{f} (\lnot \hbox {K}_{\pi i}\, \& \, \hbox {K}_{\pi j})\). Given that ‘\(\lnot \)’ is the symbol in the DRS/SDRS language that corresponds to natural-language negation, this semantic rule invites conflation of negation and the illocutionary consequences of denial—what I have dubbed no-gation.
This characterization of nul-R( \(\pi _{{i}}, \pi _{j}\) ) will suffice for my purposes here. But a more sophisticated semantics would have to be articulated to account for dialogues in which denials are themselves the targets of denials. Such second-order denials will give rise to multiply-nullified relational conditions. As Asher and Lascarides correctly observe (2003, p. 347), a doubly-nullified r-conditions nul-nul-\(R(\pi _{{i}}, \pi _{j})\) must be semantically equivalent to \(R(\pi _{{i}}, \pi _{j})\). Thus, in general, if a relational condition \(R(\pi _{j}, \pi _{k})\) is nullified an odd number of times, it is equivalent to nul-R( \(\pi _{{i}}, \pi _{j}\) ), and if it is nullified an even number of times, it is equivalent to \(R(\pi _{{i}}, \pi _{j})\).
Asher and Lascarides’ (1998a, b, 2003) treatment of presupposition in SDRT is too complicated for me to address in detail here. In this note I will mention several significant differences, and briefly indicate why I favor my approach. One significant difference is that Asher and Lascarides repudiate any process of presupposition accommodation. This is problematic, because they also claim to endorse van der Sandt’s conception of presupposition triggers as anaphoric expressions, and moreover, they do not allow for initial backgrounds to be anything other than the empty SDRS. This combination of commitments makes it difficult for Asher and Lascarides to account for discourse initial utterances that contain presupposition triggers: If the background SDRS is empty, and there is no process of accommodation, how can such presuppositions be resolved? How, for example, do Asher and Lascarides analyze a discourse-initial utterance of ‘The king of France is bald’? (I am going to ignore the presupposition triggered by ‘France’.) Asher and Lascarides attempt to solve this problem, but it is not clear that their proposed solution is coherent. They claim that utterances containing presupposition triggers in effect encode two SDRSs, one corresponding to “asserted content” and the other corresponding to “presupposed content.” They also claim—seemingly inspired by the anaphoric conception—that “the grammar produces a representation of presupposed information which is always incomplete or under-specified” (1998a, p. 252). Asserted information, in contrast, is not grammatically determined to be under-specified. Thus, according to the proposed solution, a discourse initial utterance of ‘The king of France is bald’ is processed by first adding the fully-specified asserted content to the empty background, and then using this information for the subsequent specification of the under-specified presupposed information. But this gets things backwards. It is what is asserted that is under-specified; what is presupposed is the fully-specified information that must be taken as given in order to provide the requisite specifications. In terms of a discourse initial utterance of ‘The king of France is bald’, the utterance presupposes the fully-specified information that there is a king of France, and it asserts the under-specified information that he is bald. Indeed, this is precisely how Asher and Lascarides (1998a, p. 272) informally characterize the presupposed and asserted contents in this example. Because of this confusion concerning what it is that is under-specified it is difficult to discern a coherent theoretical motivation for Asher and Lascarides’ formal apparatus. Another significant difference may be specific to the case of definite NPs. On Asher and Lascarides’s treatment “a presupposition” is like a mini-assertion, and thus Asher and Lascarides are compelled to maintain that “presupposed content” must be rhetorically related to the background, just as “asserted content” must be so related. (Of course they are forced to make an exception to this rule in the case of the initial assertion; here essentially the same conundrum described above resurfaces.) But this requirement is also problematic, for, at least in the case definite NPs, the requirement that “presuppositions” be rhetorically related to the background breaks the otherwise compelling connection between rhetorical relations and speech acts: What relational speech act—other than referring or presupposing—does one perform in using, say, a pronoun? For example, does use of the masculine-singular pronoun ‘he’ itself ever explain, or elaborate, or narrate or .... a previous utterance? It is no surprise then that Asher and Lascarides (1998a, pp. 251, 269) maintain that “presupposed content” is nearly always rhetorically related to the background SDRS via a very complex, and thus seemingly contrived, relation they dub, background( , ). As far as I can discern, whatever explanatory work is accomplished by Asher and Lascarides through the attachment of “presupposed content” with background( , ) can, at least in the case of definite NPs, be more simply and more plausibly accomplished by the sort of accommodation and binding procedures I present here.
Thus on my Strawson-inspired view, descriptive information associated with definite NPs is equivalent to the phi-features of pronouns.
I will utilize comment( , ) as an all-purpose bi-veridical rhetorical relation. I assume that the glue logic mandates comment( \(\pi _{{i}}, \pi _{j}\) ) if \(\hbox {K}_{\pi i}\) provides a topic for \(\hbox {K}_{\pi j}\). Thus comment( \(\pi _{{i}}, \pi _{j}\) ) is the converse of Asher and Lascarides’ (2003, p. 460) equivalent, but less intuitive, rhetorical relation \(\Downarrow (\pi _{j}, \pi _{{i}})\).
Asher and Lascarides (1998a, b, 2003) allow that the resolution of one of these under-specifications can influence the way the other is resolved, and they moreover allow that the resolution may occur in either order. Thus, how the presupposition is resolved might affect how \(R_{?}\)( , ) and \(\pi _{?}\) are specified, and vice versa. This more holistic approach is compatible with my simplistic version of SDRT, but here I will assume that \(R_{?}\)( , ) and \(\pi _{?}\) are specified first, and the presuppositions resolved second.
I am going to set aside cases where the presupposition triggered by ‘Harry’ are also accommodated. Of course such accommodations are possible.
Note that this SDRS is semantically equivalent to the SDRS that would result from an utterance of ‘Harry has a wife; she is rich’.
Asher and Lascarides (1998a, b, 2003) do not countenance a rhetorical relation as general as saying. Here I assume, again for the sake of simplicity, that the said component of \((\hbox {A}_{1})\) stands in the straightforward bi-veridical relation comment( , ) to a segment of the background. More realistic exchanges would require more complex rhetorical relations.
The definition of availability provided by Asher and Lascarides (2003, p. 149) further requires that R( , ) not be a “structural relation” such as contrast( , ). For the sake of simplicity I ignore this additional restriction here.
The most significant difference is that Asher and Lascarides (2003) and Lascarides and Asher (2009) maintain that correction( \(\pi _{{i}}, \pi _{j}\) ) can hold only if “there is a bijection \(\upzeta \) from the focus background structure of [\(\hbox {K}_{j}\)] onto the logical forms of subclausal constituents of [\(\hbox {K}_{{i}}\)]” (2003, p. 351), where \(\upzeta \) in effect divides \(\hbox {K}_{{i}}\) into a part that is to be retracted, and a part that is to remain in the background. But this alleged bijection requirement seems to be violated by many felicitous corrections. Indeed, presupposition corrections such as \((\hbox {A}_{3})\) seem to be obvious counterexamples to the requirement. The information structure of \((\hbox {A}_{3})\) ‘Harry is a bachelor’ does not seem to divide the content of its target \((\hbox {A}_{1})\) ‘Harry’s wife is rich’ into a part that is to be retracted, and a part that is to remain in the background; in the case of such direct presupposition denials, it seems that all the illocutionary effects of the target must be retracted. Given this, it is not surprising that when Asher and Lascaredes (2003, p. 359) briefly discuss a direct presupposition denial involving a definite NP, the bijection requirement is entirely ignored. The problem is briefly touched upon in Lascarides and Asher (2009, 22), where it suggested that in such cases “arguably the focus [of the correction] is now the whole sentence.” The suggestion seems to be that in these cases the background part of the “bijection” is empty, as is the corresponding part of the target, the part that is to remain in the background. I think a more plausible response to the counterexamples is simply to reject the bijection requirement.
Asher and Lascarides (2003) refer to the exemplified speech act as correction. But I prefer the hyphenated term ‘denial-correction’ to represent that two aspects of this complex speech-act. Note that it is possible to deny a target assertion without correcting it. For example, I might respond to an assertion by simply uttering “no.” In so doing, I deny the target assertion, but I do not correct it—I do not express what I take to be the problem with the target assertion.
In terms of Asher and Lascarides (2003) formalization of the glue logic, I am proposing that it contains the following axiom for denial-correction( , ):
-
(\(?(\upalpha , \upbeta , \uplambda ) \wedge \hbox {K}_{\upbeta }\) is inconsistent with \(\hbox {K}_{\upalpha }) > \) denial-correction(\(\upalpha , \upbeta , \uplambda ))\)
And in order to coerce the pragmatic enrichment of denials and/or their target utterances, the glue logic must also contain the following “meaning postulate”:
-
Denial-Correction(\(\upalpha , \upbeta \)) only if \(\hbox {K}_{\upbeta } \) is inconsistent with \(\hbox {K}_{\upalpha } \)
-
Asher and Lascarides use the term ‘disputed’ where I use ‘nullified’. I think their use of the term ‘disputed’ is misleading. One of the illocutionary effects of denial-correction is the retraction of the illocutionary effects of the target assertion. If the speaker of the target assertion accepts the denial-correction of her assertion, then she accepts that the illocutionary effects of her assertion have been, at least for the purposes of the conversation, retracted. But if both speakers accept that the asserted content is retracted, it is misleading to continue to characterize this content as disputed. Categorizing the illocutionary effects of denial as “putting in dispute” the illocutionary effects of the target assertion conflates the theoretical question of the illocutionary effects of denial with the question of whether or not, in a particular conversation, the denial is accepted by all participants. For proposals concerning how disputed information is distinguished from settled information in divergent dialogues, see Asher and Gillies (2003), Asher and Lascarides (2003) and Lascarides and Asher (2009).
Rule (7) happens to yield intuitively correct predictions when applied to nul-denial-correction; such nullified denial-corrections will arise if a denial is itself denied. But rule (7) will not generally yield intuitively correct predictions when applied to r-conditions that are not (at least) left-veridical. Additional rules would have to be provided to account for the nullification of non-left-veridical rhetorical relations, e.g., consequence( , ).
The analyses of van der Sandt (1991, 2003) and Van der Sandt and Meier (2003) are inadequate because they attempt to explain denials solely by appeal to the non-monotonic illocutionary effects of denial, while ignoring the monotonic effects of negated echoic corrections. According to Van der Sandt and Meier (2003) the sole illocutionary effect of an echoic correction \(({\upsigma }_{2})\) is to commit the speaker to rejection of some of the information communicated by the target assertion \(({\upsigma }_{1})\); crucially, their mechanism of directed reverse anaphora does not commit the speaker to the content of their negated echoic utterance. Thus, for example, on van der Sandt and Maier’s analysis, if Mary utters ‘Obama is not a socialist’ as an echoic correction of a previous assertion of ‘Obama is a socialist’, she does not thereby add to the background context that Obama is not a socialist. In the case of presupposition denials this aspect of van der Sandt and Maier’s view seems appropriate: If Mary utters ‘the king of France is not bald’ as an echoic presupposition correction, then, given that there is no king of France, it is not clear what proposition she could be adding to the common ground. But in the cases of proposition and implicature denials van der Sandt and Maier’s view that echoic denials are merely echoic seems to be incorrect. It is telling that neither Van der Sandt (1991), nor Van der Sandt (2003), nor Van der Sandt and Meier (2003) considers a straightforward example of a proposition denial.
Note that u and v in \(\hbox {U}(\hbox {K}_{\pi 1})\) are available specifications for \(d_{1?}\) and \(d_{2?}\), even though \(\hbox {K}_{\pi 1}\) is no-gated as a result of the denial \((\hbox {H}_{2})\). This is because u and v are not introduced in \(\hbox {K}_{\pi 1}\); rather u and v are introduced in \(\hbox {K}_{\pi 0}\).
Asher and Lascarides (1998a, 2003) invoke the requirements on rhetorical relations imposed by the glue logic to account for many other felt implicatures. Moreover, though I cannot pursue the point here, the process of denial-correction( , ) induced pragmatic enrichment also explains the felicity of what Horn (1985, 1989 classified as instances of “metalinguistic negation.” The relationship between “metalinguistic negation” and presupposition denials will be briefly discussed in the concluding section.
I concede that have not quite accounted for this last aspect of the intuitive truth-conditions; i.e. I have not accounted for the intuition that exchange (H) entails that Harry’s wife being an anthropologist explains why she is not rich. (Note that it is natural to explicitly connect \((\hbox {H}_{2})\) and \((\hbox {H}_{3})\) with the cue word ‘because’: ‘Harry’s wife is not rich, because she’s an anthropologist’.) I omit this aspect of SDRT because it significantly complicates the semantics, yet is not central to my concerns here. For my purposes here, it suffices to think of this additional entailment as being an aspect of pragmatically enriched \(\hbox {K}_{\pi 3}\).
Note that an “intermediate accommodation” of the presupposition into \(\hbox {K}_{\pi 1^{*}}\) does not make sense: we eavesdroppers cannot understand Mary’s denial is somehow requiring John’s target assertion to be exploiting the repair procedure of accommodation.
I here ignore that the glue logic prohibits these specifications because \(\hbox {K}_{\pi 1}\) is not incompatible with \(\hbox {K}_{\pi 2}\): speaking loosely, \(\hbox {K}_{\pi 1} \)is true if and only if Harry has a rich wife, and \(\hbox {K}_{\pi 1}\) is true if and only if Harry has a (perhaps distinct!) wife who is not rich. Ordinarily the requirements imposed by the glue logic for these specifications might be met by the pragmatic enrichment of \(\hbox {K}_{\pi 2}\) to produce an enriched SDRS \(\hbox {K}_{\pi 2+}\) that is incompatible with target SDRS \(\hbox {K}_{\pi 1}\), but in this case such pragmatic enrichment will not succeed in producing an adequate interpretation of exchange (I) in its entirety. The problem is that, even overlooking the violation of the inconsistency felicity requirement, the follow-up-correction \((\hbox {I}_{3})\) will result in an inconsistent SDRS.
I here again ignore the harmless redundancy of conditions \((\hbox {ii}_{\mathrm{a}})\) and \((\hbox {ii}_{\mathrm{b}})\).
SDRS \((\hbox {BG}_{\mathrm{A}}^{2})\) is identical to SDRS \((\hbox {BG}_{\mathrm{H}}^{2})\).
That the semantics provides a solution to the denial problem was already apparent in the treatment of exchange (H) presented in Sect. 4.3.
It is possible to attach \(\pi _{2}\) to \(\pi _{1a}\), and for some continuations of the discourse such interpretation would be correct. That is, some discourses require that we eavesdroppers interpret Mary as accepting the presupposition that \((\hbox {A}_{2})\) requires her to accommodate, but denying the asserted component of \((\hbox {A}_{2})\). There is thus an alternative version of the participant-accommodation interpretation under which Mary accepts what John has presupposed, but she does not accept what John asserts. For exchange (A), however, the preferred attachment point is (still) \(\pi _{1}\).
That the marked feel of presupposition denials is at least in part explained by the process of coerced reinterpretation is explicit in Burton-Roberts (1989a, b) and Carston (1998, 2002), and implied by the opening sentence of Horn (1985): “So-called ‘external’ or ‘marked’ negation is often exemplified by the reading of The king of France is not bald which is forced by the continuation ... because there is no king of France” (1985, 121).
Carston thus differs from Burton-Roberts and Horn in that she denies that presupposition denials require any special non-truth-functional use of negation. Carston, appropriately in my view, finds Horn’s notion of a pragmatic ambiguity of negation suspect (2002, p. 296).
References
Asher, N. (1993). Reference to abstract objects in discourse. Dordrecht: Kluwer Academic Publishers.
Asher, N. (1995). From discourse macro-structure to micro-stricture and back again. In H. Kamp & B. Partee (Eds.), Proceedings of the conference on semantics in context SFB340 Report. Stuttgart: University of Stuttgart.
Asher, N., & Gillies, A. (2003). Common ground, corrections, and coordination. Argumentation, 17, 481–512.
Asher, N., & Lascarides, A. (1998a). The semantics and pragmatics of presupposition. Journal of Semantics, 15, 239–299.
Asher, N., & Lascarides, A. (1998b). Bridging. Journal of Semantics, 15, 83–113.
Asher, N., & Lascarides, A. (2003). The logics of conversation. Cambridge: Cambridge University Press.
Asher, N., & Lascarides, A. (2011). Reasoning dynamically about what one says. Synthese, 183, 5–31.
Beaver, D. (2001). Presupposition and assertion in dynamic semantics. Stanford: CSLI Publications.
Beaver, D., & Geurts, B. (2013). In E. Zalta (Ed.), Presupposition. In: The Stanford encyclopedia of philosophy. http://plato.stanford.edu/archives/fall2013/entries/presupposition.
Boer, S., & Lycan, W. (1976). The myth of semantic presupposition. Bloomington: Indiana University Linguistics Club.
Burton-Roberts, N. (1989a). On Horn’s dilemma: Presupposition and negation. Journal of Linguistics, 25(1), 95–125.
Burton-Roberts, N. (1989b). The limits to debate. Cambridge: Cambridge University Press.
Carston, R. (1998). Negation, “presupposition,” and the semantics-pragmatics distinction. Journal of Linguistics, 34, 309–350.
Carston, R. (1999). Negation, ‘presupposition’ and metarepresentation: A response to Burton-Roberts. Journal of Linguistics, 35, 365–389.
Carston, R. (2002). Thoughts and utterances. Oxford: Blackwell.
Chemla, E., & Schlenker, P. (2012). Incremental vs. symmetric accounts of presupposition projection: An experimental approach. Natural Language Semantics, 20(2), 177–226.
Cummins, C., Amaral, P., & Katsos, N. (2013). Backgrounding and accommodation of Presuppositions: An experimental approach. In: E. Chemla, V. Homer, & G. Winterstein (Eds.), Proceedings of Sinn and Bedeutung, 17 (pp. 201–218).
Geurts, B. (1998). The mechanisms of denial. Language, 74(2), 274–307.
Geurts, B. (1999). Presupposition and pronouns. Amsterdam: Elsevier.
Geurts, B. (2007). Existential import. In I. Comorovski & K. von Heusinger (Eds.), Existence: Syntax and semantics (pp. 253–271). Dordrecht: Springer.
Hayashi, M., Raymond, G., & Sidnell, J. (2013). Conversation repair and human understanding: An introduction. In M. Hayashi, G. Raymond, & J. Sidnell (Eds.), Conversation repair and human understanding (pp. 1–40). Cambridge: Cambridge University Press.
Horn, L. (1985). Metalinguistic negation and pragmatic ambiguity. Language, 61, 121–174.
Horn, L. (1989). A natural history of negation. Chicago: Chicago University Press.
Kamp, H., & Reyle, U. (1993). From discourse to logic. Dordrecht: Kluwer.
Kempson, R. (1975). Presupposition and the delimitation of semantics. Cambridge: Cambridge University Press.
Lascarides, A., & Asher, N. (2009). Agreement, disputes and commitments in dialogue. Journal of Semantics, 26(2), 109–158.
Roberts, C., Simons, M., Beaver, D., & Tonhauser, J. (2009). Presupposition, conventional implicature, and beyond: A unified account of projection. In N. Klinedist & D. Rothschild (Eds.), Proceedings of the workshop on new directions in the theory of presuppositions. Toulouse: ESSLI.
Russell, B. (1905). On denoting. Mind, 14, 479–493.
Russell, B. (1919). Introduction to mathematical philosophy. London: George Allen and Irwin Ltd.
Sacks, H., Schegloff, E., & Jefferson, G. (1974). A simplest systematics for the organization of turn taking for conversation. Language, 50(4), 696–735.
Schegloff, E., Jefferson, G., & Sacks, H. (1977). The preference for self-correction in the organization of repair in conversation. Language, 53(2), 361–382.
Schlenker, P. (2011). DRT with local contexts. Natural Language Semantics, 19(4), 373–392.
Schouebye, A. (2009). Descriptions, truth value intuitions, and questions. Linguistics and Philosophy, 32(6), 583–617.
Stalnaker, R. (1974). Pragmatic presuppositoins. In M. Munitz & P. Unger (Eds.), Semantics and philosphy (pp. 197–214). New York: New York University Press.
Stalnaker, R. (1979). Assertion. In P. Cole (Ed.), Syntax and semantics (Vol. 9). London: Academic Press.
Strawson, P. (1950). On referring. Mind, 59, 320–344.
Strawson, P. (1964). Identifying reference and truth values. Theoria, 30(2), 96–118.
Tonhauser, J., Beaver, D., Roberts, C., & Simons, M. (2013). Towards a taxonomy of projective content. Language, 89(1), 66–109.
Van der Sandt, R. (1987). Presupposition and discourse structure. In Proceedings of the 6th Amserdam Colloquium, University of Amsterdam (pp. 312–330). (Reprinted with revisions in Semantics and contextual expression, pp. 267–294, by R. Bartsch, J. van Benthem, & O. van Emde Boas, Eds., 1989, Dordrecht: Foris.)
Van der Sandt, R. (1991). Denial. Chicago Linguistic Society, 27(2), 331–344.
Van der Sandt, R. (1992). Presupposition projection as anaphora resolution. Journal of Semantics, 9, 333–377.
Van der Sandt, R. (2003). Denial and presupposition. In P. Kühnlrin, H. Rieser, & H. Zeevat (Eds.), Perspectives on dialogue in the new millennium (pp. 59–77). Amsterdam: John Benjamins.
Van der Sandt, R., & Meier, E. (2003). Denial and correction in layered DRT. In Proceedings of DiaBruck.
Van Leusen, N. (2004). Incompatibility in context: A diagnosis of correction. Journal of Semantics, 21, 415–441.
Von Fintel, K. (2004). Would you believe it? The King of France is back! Presuppositions and truth-value intuitions. In M. Reimer & A. Bezuedenhout (Eds.), Descriptions and beyond. Oxford: Oxford University Press.
Wilson, D. (1975). Presupposition and non-truth-conditional semantics. London: Academic Press.
Yablo, S. (2006). Non-catastrophic presupposition failure. In J. J. Thomson & A. Byrne (Eds.), Context and modality: Themes from the philosophy of Robert Stalnaker. Oxford: Oxford University Press.
Acknowledgments
I presented preliminary versions of this paper at the following conferences and colloquia: the Workshop on Language and Interpretation held at UNAM, March 20–23, 2013; the University of Kentucky Philosophy Colloquium and Linguistics Seminar Series, October 4, 2013; the University of Iowa Philosophy Colloquium, November 15, 2013; the Phling Workgroup at Northwestern University, November 22, 2013; and the meeting of the Pacific Division of the APA, San Francisco CA, April 16–20, 2014. I am grateful to all the participants in these events who pushed me to clarify my views, but in particular I would like thank Robin Carston, Bart Geurts, Tim Sundell, Michael Glanzberg, Benjamin Rohrs and Dylan Sabo for their helpful and challenging feedback. Finally, this paper has been revised substantially in response to comments from several anonymous referees for Synthese. The paper benefited from their detailed criticisms, and I am grateful for their assiduous service to the profession.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Clapp, L. On denying presuppositions. Synthese 194, 1841–1900 (2017). https://doi.org/10.1007/s11229-016-1026-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11229-016-1026-z