Everything that rises must converge: Toward a unified account of inquisitive and assertive rising declaratives

Chapter 12 Everything that rises must converge: Toward a unified account of inquisitive and assertive rising declaratives Daniel Goodhue Leibniz-Zentrum Allgemeine Sprachwissenschaft (ZAS) In English, matrix declaratives with a final rising intonation typical of polar questions are frequently used as a kind of biased question: they can only be used when there is contextual evidence in favor of the proposition denoted by the declarative. However, not all rising declaratives are used to pose a question about their content – some are used to assert the content of the declarative, while raising a second issue. In this paper, I offer a unified account of rising declaratives that seeks to explain both of these kinds of uses while positing unitary meanings for clause types and intonations. This goal cannot be achieved if we take the illocutionary force of an utterance to be completely determined by clause type and linguistic intonation, as many recent accounts have done. Instead, I propose that clause type and intonation merely constrain what a speaker could intend to do with them; pragmatic inference must play a key role in enabling an audience to uncover the speaker’s illocutionary intention. In other words, there can be no hard and fast conventional discourse effects tied to particular clause type +intonation pairings. I demonstrate that the proposed account enables a derivation of assertive force, and comparisons to other recent accounts are made. 1 Introduction The main claim of this paper is that one particular rising intonation, the polar question rise, has only one specific meaning across its disparate uses, roughly, the speaker does not commit to a relevant proposition. My account unifies inquisitive Daniel Goodhue. 2025. Everything that rises must converge: Toward a unified account of inquisitive and assertive rising declaratives. In Tue Trinh, Anton Benz, Daniel Goodhue, Kazuko Yatsushiro & Manfred Krifka (eds.), Biased questions: Experimental results and theoretical modelling, 371–401. Berlin: Language Science Press. DOI: 10.5281/zenodo.17158198 Daniel Goodhue and assertive rising declaratives by deriving their distinct global interpretations from the same inputs – polar question rises and declarative clauses – used in different contexts.1 The following classic examples demonstrate that rising declaratives (RDs) can be used to ask biased questions. The questions are biased in the sense that they require contextual evidence in favor of the proposition denoted by the declarative. Throughout this paper, I use ‘↗’ (pronounced “rise”) to represent the utterance final rising intonation typical of polar questions in English. The relevant intonation will be discussed further in Section 2. (1) S is in her windowless office. A has just arrived holding a wet umbrella and raincoat. a. S: Hey! It’s raining↗ b. S: Hey! Is it raining↗(based on Gunlogson 2003: 96) Intuitively, both the RD in (1a) and the polar interrogative in (1b) are felicitous means of asking whether it is raining in the context of (1). Contrast this with (2): (2) S is in her windowless office. A has just arrived, and exhibits no evidence whatsoever about the weather outside. a. S: # Hey! It’s raining↗ b. S: Hey! Is it raining↗(based on Gunlogson 2003: 95) The context of (2) lacks any evidence for rain. Intuitively, the RD in (2a) is infelicitous, while the polar interrogative in (2b) is just fine. In prior work, examples like these establish (3) as a robust generalization about RDs (see e.g. Beun 2000, Gunlogson 2003, 2008, Truckenbrodt 2006, 2009, 2012, Trinh & Crnič 2011, Malamud & Stephenson 2015, Farkas & Roelofsen 2017, Krifka 2017, Westera 2017, 2018, Jeong 2018, Rudin 2018, 2022). (3) (Inquisitive) Rising Declaratives are felicitous only if there is contextual evidence in favor of the content of the declarative clause. What we learn from such RDs is that declaratives are not reserved for assertions. With a particular rising intonation, declaratives can be used to ask a biased question. 1So the title does not refer to every rising intonation, just polar question rises. The point is that polar question rises converge on a single meaning, even though the global interpretation of utterances they appear in may vary. 372 12 Everything that rises must converge However, biased questions are not the only use for rising declaratives. RDs can also be used to assert. Consider the RDs in (4), (5), and (6), which are used by S to assert their propositional content.2 (4) A: Do you speak Spanish? S: I speak Ladino↗ (Jeong 2018, Farkas & Roelofsen 2017, based on Ward & Hirschberg 1985) (5) A: What are you eating? S: This is a persimmon↗ (6) S isn’t sure that he is in the right doctor’s office. He says to the receptionist: S: My name is Mark Liberman↗(Pierrehumbert 1980: 62, from Mark Liberman p.c.) Besides being used to assert their content, the RDs in (4), (5), and (6) also seem to raise a second issue. In (4) this is something like Is Ladino close enough to Spanish for your purposes?, while in (5) and (6) this is something like Is that enough information? or Have you heard of persimmons/me before?. Most of the literature on RDs cited above has either ignored assertive RDs, set them aside, or tried to account for them separately from inquisitive RDs.3This has usually been justified by the claim that there are two distinct rising contours, one used in inquisitive RDs, the other in assertive RDs. The idea is that each contour makes a distinct meaning contribution resulting in the distinct inquisitive and assertive illocutionary forces observed. In Section 2, I will argue that the evidence for this view is weak, and provide further evidence that speaks against it. As a result of this empirical evidence, as well as for reasons of theoretical parsimony, I will argue that there is only one relevant rising intonation, ↗, with a single meaning attached to it that can explain its use in both inquisitive and assertive RDs, as well as the fact that it is used in most matrix polar interrogatives (see Hedberg et al. 2017 for corpus evidence that over 90% of American English polar interrogatives rise utterance finally, pace Geluykens 1988). Furthermore, I 2Ward & Hirschberg (1985) introduced (4) as an example of the rise-fall-rise contour (L*+H LH%), which is distinct from the ↗contour used in RDs. Nevertheless, ↗is also felicitous in (4), though probably not preferred. On the other hand, (5), (6), and other examples of assertive RDs below are not felicitous with rise-fall-rise. 3I believe that assertive RDs are likely related to ‘uptalk’ and ‘high rising terminals’. If so, then they have also been discussed in the sociolinguistics literature (e.g. McLemore 1991, Fletcher et al. 2005, Ladd 2008, Shokeir 2008: a.o.). I leave a full exploration of this connection to future work. 373 Daniel Goodhue will adopt a semantics for clause types in which declaratives denote propositions while interrogatives denote sets of propositions (answer sets). But if there is a single meaning for ↗, and a single denotation for declaratives, then clearly the combination of these two components alone cannot completely explain the variation in speech act interpretation we see across inquisitive and assertive rising declaratives. In other words, intonation plus clause type does not always determine illocutionary force. This is contrary to the predictions of prior work such as Farkas & Roelofsen (2017), Jeong (2018), and Rudin (2018), which claim that an utterance of a specific clause type with a specific intonation results in exactly one discourse update effect (though the details differ substantially across these accounts). The solution I will propose is to abandon the view that clause type-intonation pairings are specified with conventional discourse effects, and instead allow pragmatic inference to play a greater role in enabling the audience to uncover the illocutionary force intended by the speaker of an utterance. In other words, in order to simultaneously account for inquisitive and assertive rising declaratives, the contributions to illocutionary force made by clause type and intonation need to be weaker than hypothesized in the recent literature on rising declaratives. In Section 2, I discuss the intonational facts to motivate a unified account of inquisitive and assertive RDs. Then I lay out the ingredients of the unified account in Section 3. In Section 4, I apply the account to the basic data, as well as to incredulous uses of RDs, and I briefly discuss its potential application to rising imperatives. In Section 5, I show how the account enables a derivation of assertive force. Finally, I briefly compare the account to prior work in Section 6.1, and raise issues for future work in Section 6.2. 2 Perspectives on the intonation of RDs Much work on rising declaratives has claimed that they come with two distinct intonational (phonological) contours, one for inquisitive RDs and another for assertive RDs, with each contour playing a crucial role in determining the illocutionary force of the RD. If this is correct, it means that the two kinds of RDs are orthogonal and can be given independent analyses. In this section, I will cast doubt on this view, and show that the empirical facts are consistent with at least three other views in which intonation does not neatly distinguish the illocutionary force of RDs. If one of these latter views is correct, it means that the two kinds of RDs are not orthogonal and must be given a unified analysis. I will then develop such an analysis in the remainder of the paper. 374 12 Everything that rises must converge The earliest empirical evidence motivating the hypothesized intonational distinction between assertive and inquisitive rising declaratives was based on researcher judgments (Pierrehumbert 1980, Pierrehumbert & Hirschberg 1990, Hirschberg & Ward 1995). The hypothesis is that assertive RDs have a high rising contour, represented by H* H-H%, while the standard polar question contour in polar questions and inquisitive RDs is low rising, L* H-H%.4It’s clear from the discussions in these references that the relevant perceptual distinction is meant to be in the height of the nuclear pitch accent, with L* at the bottom of the speaker’s range, while H* is in the middle. Truckenbrodt (2012) builds on ideas in Pierrehumbert & Hirschberg (1990), Hirschberg & Ward (1995), and Bartels (1999). He analyzes H* as signaling addition of a salient proposition to the common ground, while Hsignals the questioning of a salient proposition (L* and Lare treated as meaningless defaults). Thus L* H-H% questions the proposition uttered, while H* H-H% asserts the proposition uttered while questioning a related salient proposition.5,6 There is some reason to doubt this empirical picture in which L* H-H% correlates with inquisitive RDs while H* H-H% correlates with assertive RDs. First, Hedberg et al. (2017) claim that their corpus data shows that both contours are used in matrix polar interrogatives. Since matrix polar interrogatives only have an inquisitive interpretation, the meaning contribution of H* H-H% would need to be canceled somehow. Second, Nilsenovà (2006) uses corpus data for stimuli in a comprehension experiment that shows that L* H-H% and H* H-H%, as well as L* L-H%, all significantly increase inquisitive interpretations of declarative clauses. This shouldn’t happen if H* H-H% is the assertive contour for RDs. These results 4A brief primer on the Tones and Break Indices system (ToBI; Veilleux et al. 2006): There are high (H) and low (L) tones. ‘T*’ indicates a pitch accent (a tone on a syllable that is more perceptually salient/stressed), ‘T%’ an intonational boundary (the end of an intonational unit, usually the end of a sentence), and ‘T-’ a phrase tone leading up to the boundary tone. The final pitch accent in a sentence is called the nuclear pitch accent, and the intonation from that point on is the nuclear contour. The nuclear pitch accent is the most salient intonational stress, even though it usually isn’t the greatest pitch maximum/minimum acoustically-speaking. 5One mystery for uses of H* H-H% in Truckenbrodt’s analysis is why the H* always targets the content of the utterance while Htargets some other salient proposition. The account predicts these roles to be reversible, contrary to fact. My analysis in Section 3 pursues the idea that ↗ can target salient propositions, but derives which propositions are targeted from pragmatics. 6Interestingly, Pierrehumbert (1980: 62–63) does not claim that assertive RDs show that an intonational distinction has a strong grammatical link to speech act meaning. On the contrary, despite distinguishing H* H-H% from L* H-H%, she takes them to be family members of a single yes/no question intonation, and her point is that assertive RDs like (6) show that this yes/no question intonation does not force the utterance to be interpreted as a yes/no question about the truth of the proposition. She then briefly sketches an analysis of intonational meaning that is actually in the same spirit as the one I pursue in Section 3 through Section 5. 375 Daniel Goodhue call into question the view that the inquisitive/assertive split in RDs correlates with L* H-H% and H* H-H% respectively. More recently, a view has emerged in which the inquisitive/assertive split is claimed to correlate with the height of the final H% boundary tone: Inquisitive RDs are claimed to rise more steeply to a higher final boundary tone than the contour associated with assertive RDs (Jeong 2018, Rudin 2018, Westera 2018).7 Jeong (2018: 320ff.) reports on a series of multi-participant comprehension experiments in which the intonation of RDs is manipulated via prosodic resynthesis so that the height of the pitch accent remains constant (neither high nor low, but in the middle of the speaker’s range), while the boundary tone is higher in some conditions than others. The results show that the steeper the rise is, the more likely participants are to arrive at an inquisitive interpretation (absent other contextual factors that make inquisitive or assertive interpretations more likely); for the shallowest rise, participants choose between inquisitive and assertive interpretations at around chance levels.8Jeong argues that these results can be explained if there are two intonational contours with meanings that determine the illocutionary force of RDs: a steeper one that leads to inquisitive force, and a shallower one that leads to assertive force. I turn now to showing that there is a competing analysis available (that comes in a few sub-flavors) in which intonation correlates, not with the split between inquisitive and assertive RDs, but instead with a split between incredulous and non-incredulous RDs. To appreciate this, consider the following taxonomy of RDs (cf. the similar taxonomy in Jeong 2018). (7) Taxonomy of Rising Declaratives Inquisitive RDs Incredulous RDs Confirmative RDs Assertive RDs 7Jeong (2018) and Rudin (2018) continue to use the L* H-H% and H* H-H% ToBI transcriptions to distinguish these two contours, despite that the key phonetic distinction is in the height of final boundary tones rather than nuclear pitch accents. That’s because the phonological distinction between L* H-H% and H* H-H% could in principle result in an observed phonetic distinction in the height of the final H% boundary tone. Thus the phonological and interpretational claims of Pierrehumbert & Hirschberg (1990), Hirschberg & Ward (1995) and Truckenbrodt (2012) could still be viable, even if the claim that the perceptual distinction is in the height of the pitch accent may not be. Thanks to Sunwoo Jeong, Jeffrey Lidz, and Michael Wagner for discussion on this point. 8Jeong (2018) also shows that relative speaker/addressee knowledgeability strongly impacts interpretation, largely taking precedence over intonation. This factor will play an important role in my analysis as well. 376 12 Everything that rises must converge According to (7), uses of inquisitive RDs can be sorted into two kinds: incredulous uses and confirmative uses. (8–10) provide a minimal triple demonstrating the taxonomy in (7). In (8), S utters the RD incredulously because they are shocked by A’s claim that the girl is only nine. (8) Incredulous S and A are watching a girl give a very professional performance in a school debate. S thinks that she must be at least 13 years old. A: I can’t believe she’s only 9. S: She’s nine↗ In (9), S utters the RD confirmatively; they are not shocked that the girl is nine, they just want to double check that fact. (9) Confirmative S and A are buying a birthday card for the daughter of A’s friend. While searching for a card for the correct age, S thinks A told him previously that the girl has just turned nine, but he wants to confirm it. S: She’s nine↗ In (10), S asserts that their daughter is nine, but doesn’t know whether there is still room for kids in her age group. (10) Assertive S wants to enroll his daughter in music lessons with A. S: My daughter wants to study tuba. A: Okay, but there are limited places for each age group, and some age groups have already filled up. How old is she? S: She’s nine↗ I will now lay out three possible analyses of the intonational distinctions to be found among the kinds of RDs in (7), and then show how each analysis can explain the experimental results in Jeong (2018) without positing two phonological intonations that correlate with force in the inquisitive/assertive split. Starting with the distinction in the height of the final boundary tone postulated by Jeong (2018), Rudin (2018), and Westera (2018), one possible analysis is that it’s a paralinguistic distinction, such that all three sub-kinds of RDs in (7) have the same phonological intonation, best transcribed in ToBI as L* H-H%, and incredulous RDs have a higher boundary tone due to increased emotional activation, as has been discussed by Gussenhoven (2004), Bänziger & Scherer (2005), Crespo Sendra et al. (2013), Westera (2017), and Goodhue (2021). 377 Daniel Goodhue Another possible analysis is that there is a phonological distinction between incredulous and non-incredulous RDs. On this view, incredulous RDs would be phonologically specified for a very high H% boundary tone, and would be produced with at least two rises, one earlier in the sentence, and another for the nuclear contour (transcribed in ToBI as L* HL* H-H%). In (8), this would mean a low pitch accent followed by a rise on she’s, and then a fall back to a low nuclear pitch accent on nine that rises to a very high boundary tone. Intuitively, such a double rise is felicitous in (8), but infelicitous in both (9) and (10). In the production study of Goodhue et al. (2016), incredulity contexts elicited double rises to very high boundary tones in the majority of trials.9,10 Finally, the height of the final boundary tone may correlate inversely with the speaker’s certainty level about the proposition 𝑝expressed: the higher the tone, the less certain. The speaker of a confirmative RD tends to think 𝑝is likely (the basis for the choice to confirm 𝑝as opposed to some other proposition), which results in a lower H% boundary tone. The speaker of an incredulous RD takes 𝑝 to be unlikely (hence the incredulity), resulting in a higher H% tone (Section 4.2 will nuance this view of the speaker’s possible stances toward 𝑝in incredulous RDs). Each of these views of possible intonational distinctions among RDs can explain Jeong’s (2018: 320ff.) experimental results, which revealed that the steeper the rise is, the more likely participants are to arrive at an inquisitive interpretation, and in the condition with the shallowest rise, participants chose between inquisitive and assertive force at chance levels. Assume the paralinguistic view: Steeper rises imply emotional activation, and emotional activation correlates with incredulity. Furthermore, incredulous RDs are a sub-kind of inquisitive RDs in the taxonomy in (7). Thus, when participants hear a steeper rise, they are more likely to infer an inquisitive interpretation. Shallower rises on the other hand are less emotionally activated, and thus non-incredulous, which is consistent with either an inquisitive (confirmative) interpretation or an assertive interpretation. 9The height of the final boundary tone is in principle separable from the phonological contour. So a third analytic option would be that there are two linguistic contours – the normal polar question rise and an incredulous double rise – and the incredulous double rise generally has higher final H% boundary tones for paralinguistic reasons, namely because speakers using them are emotionally activated. 10The incredulous double rise discussed here is not the incredulity contour of Hirschberg & Ward (1992), which they describe as L*+H L-H%, identical to the rise-fall-rise contour but with a larger pitch excursion (Barnes et al. 2012 argue it is better transcribed as L+H* L-H%). Crucially, the incredulous double rise and the incredulity contour are perceptually distinct, and likely also have subtly different felicity conditions, which once understood should explain why the incredulity condition of Goodhue et al. (2016) elicited incredulous double rises in over 95% of trials, but only a single incredulity contour. 378 12 Everything that rises must converge Now suppose the linguistic incredulous double rise view is correct: The steeper rises were interpreted as corresponding (albeit imperfectly) to a linguistic contour specified with an incredulity meaning. Since incredulous RDs are a subkind of inquisitive RDs in (7), this led participants to an inquisitive interpretation. Meanwhile shallower rises were interpreted as corresponding to a normal polar question rise, which is consistent with both confirmative and assertive RD interpretations. Finally, on the view that steepness has an inverse correlation with certainty, it is clear why steeper rises would lead to more inquisitive interpretations. What remains to be explained is why the shallowest rises were interpreted as either inquisitive or assertive at chance levels. The answer is that a speaker can be relatively certain, even completely certain, about a proposition 𝑝and still ask a question about it, especially if the addressee is known to have more authority over whether 𝑝is true or not in the context. The following example from a squib by Sider (2022) demonstrates this well: (11) While S is feeding his daughters breakfast, he hears someone stirring upstairs. S’s daughters go up, and S hears talking. On a typical day in S’s house, this is when S’s wife usually rises. S’s daughters come back downstairs. S: Mom’s up↗(Sider 2022) (11) is a confirmative RD. As Sider says, S is relatively certain that the mother is awake. And yet the RD S utters is clearly not an assertion. On my view, it is a question, as evidenced by the fact that the children could provide a yes or no answer. S is able to ask this question because, even though S has very good evidence that 𝑝, the children are in a better position to know 𝑝: they were upstairs with the mother. Thus we understand why, on this third view in which shallower rises are consistent with high certainty levels, participants interpreted the shallowest rise as either inquisitive or assertive at chance levels in Jeong’s experiment. Here is the upshot of this discussion: The experimental results in Jeong (2018) are compatible with, but do not decide in favor of, the view proposed there in which there are two phonological intonations with meanings that play a direct role in producing inquisitive and assertive illocutionary forces (the illocutionary view). The results are also compatible with three other views in which intonation (whether phonological or paralinguistic) plays an indirect role: two in which intonation distinguishes incredulous RDs from non-incredulous kinds, and a third in which intonational steepness signals uncertainty about the propositional content 𝑝, with certainty about 𝑝being consistent with both inquisitive and assertive 379 Daniel Goodhue A: Okay, but there are limited places for each age group, and some age groups have already filled up. How old is she? S: She’s nine↗ I define questions and assertions as follows:18 (18) Definition of question/inquisitive: If S’s intention in uttering 𝑈is to raise an issue without settling that issue themselves, and with the expectation that A will settle that issue in reply by committing to a proposition in that issue, then 𝑈is a question. (19) Definition of assertion/assertive: If S’s intention in uttering 𝑈is to settle an issue by committing to a proposition in that issue, then 𝑈is an assertion. The solution I’ll pursue to explaining why the RD in (9) is a question while that in (10) is an assertion is to allow pragmatic inference to play a key role in enabling the audience to uncover the speaker’s intended speech act in uttering a rising declarative. Beyond this, there are some additional effects to be explained: (9) conveys some bias toward the proposition, while (10) raises a second issue. The QUD in (9) is How old is the birthday girl?. S utters a proposition that would answer it, that she is nine, which I’ll call 𝑞. According to (14), this puts {𝑞} on the table. However, according to (15) and (16), S’s use of ↗conveys that S is not committing to 𝑞. Given the support requirement in (17), since S is not committing to 𝑞and S is talking only to A, S must intend A to address the issue {𝑞} that S raised by giving an answer. So given the definition of question in (18), the account in Section 3 explains why the RD in (9) is regarded as a question. The second meaning effect to explain about inquisitive RDs like (9) is that they convey a bias in favor of the proposition denoted by the declarative. I pursue an explanation somewhat similar to those in both Rudin (2018) and Westera (2018): In choosing to use a RD that denotes only that she is nine, S raises an issue that contains only that one proposition, instead of the two propositions {that she is nine,that she is not nine} that would have been raised via a polar interrogative, Is she nine?. Given the support requirement in (17), S could not have done this if S did not have some reason to think that A was in a position to commit to this 18Possible counterexamples to (18) may include rhetorical questions and reflective questions. However it could be argued that rhetorical questions are in fact indirect assertions whose derivation transits through the usual understanding of questions in (18), while reflective questions are questions in which one talks to oneself, which fits with (18) when viewed in that light. 386 12 Everything that rises must converge particular proposition. Thus we understand why inquisitive rising declaratives necessarily convey a contextual bias as formulated in (3): the contextual evidence provides the justification for S to restrict A to the single proposition denoted by the declarative. Note that the speaker does not themselves need to be biased for the content of the RD. The above logic is met even if S is skeptical of that proposition, so long as S thinks A can and will commit to the proposition, as is the case for incredulous RDs (on which, more below). Note also that private speaker bias for the content of the declarative is not enough to meet the bias condition in (3). E.g., if S had private reasons to believe that it is raining in (2), and S knew that A was arriving from outside, but there was no publicly available contextual evidence of rain, then the RD in (2a) would still be infelicitous. It seems that, if S is going to raise an issue that contains only one proposition and ask A to settle it, then there must be contextual evidence available to explain why S would do so. Otherwise, S is required to provide A with more than one alternative. Now for (10): According to (15), ↗conveys that S is not committing to a relevant proposition 𝑞. However, unlike in (9), in (10) the default assumption that 𝑞is equivalent to the propositional content of the RD, that she is nine, is made contextually implausible by the fact that S is the clear epistemic authority with respect to that proposition. So the audience can safely assume that S’s use of ↗is not meant to convey lack of commitment to the declarative content. Furthermore, since S has not conveyed a lack of commitment to this content, and given S’s position of authority, along with the support requirement in (17) and the prior conversational context, the audience can further infer that S intends to commit to the proposition that she is nine, settling A’s question How old is your daughter? (this derivation of assertive commitment is revisited in greater detail in Section 5). Given the definition of assertion in (19), the account in Section 3 explains why the RD in (10) is regarded as an assertion. The second meaning effect to explain about assertive RDs like (9) is that they raise a second issue. This is caused by ↗, which still conveys that S is not committing to some proposition 𝑞that is relevant to the QUD. In (10), the goal is to enroll S’s daughter in music lessons with A. A has said that whether this can be achieved depends on whether there is room in her age group. S’s answer settles what her age group is, but leaves open whether there is still room in that age group. Thus, the proposition 𝑞that S expresses lack of commitment about is that there is still room in the nine-year-old group, since if this proposition were combined with the content of the declarative that she is nine, they would together form a strategy Γfor resolving the QUD, Can my daughter study tuba with you?. By working backward from the proposition asserted and the remaining issues 387 Daniel Goodhue that need to be resolved to achieve the goal, the audience can infer the proposition targeted for lack of commitment by ↗in assertive RDs. 4.2 Incredulous uses of RDs On one of the possible analyses I sketched for the intonation of incredulous rising declaratives in Section 2, they have their own unique phonological contour, the incredulous double rise, distinct from ↗. If so, then the analysis given for ↗in Section 3 won’t be required to explain the incredulous RD data. However, on another analysis I sketched, incredulous RDs feature the same linguistic contour ↗found in other RDs, and the difference is only a paralinguistic one affecting the height of the final boundary tone. In case this view turns out to be correct, the following discussion explores how the account of ↗in Section 3 can be applied to incredulous RDs. The speaker’s beliefs with respect to the content proposition 𝑝of inquisitive RDs can vary. In (1) for example, S could have no prior bias either way before being confronted with the evidence of A’s wet umbrella. And in the confirmative RDs of (9) and (11), S has a prior 𝑝bias. Incredulous RDs on the other hand all come with a prior not-𝑝speaker bias, though S’s beliefs may or may not change in the face of the contextual evidence for 𝑝that is required in order for the inquisitive RD to be felicitous. In (20), S believes that they should not apologize, so there is a prior not-𝑝speaker bias that persists through speaking time. (20) S believes that they did nothing wrong: A: You should apologize. S: I should apologize↗(based on Pierrehumbert & Hirschberg 1990: 292) Unlike (20), the context in (8) does not establish such a strong form of not-𝑝 speaker bias. There the speaker previously thought that the child was older than nine, and thus had a previous not-𝑝bias. But suppose A and S both take A to be in a better position to know the child’s age, perhaps because A works at the school and S is merely visiting. In that case S should be inclined to accept A’s claim that the child is nine, and so after A’s utterance, S should be on the way to believing 𝑝, even if S may not have completely accepted 𝑝yet as there is still some chance that A could be wrong. The prior not-𝑝bias in common across incredulous RDs plays a key role in explaining their incredulity. First, incredulous RDs are a subkind of inquisitive RDs, so according to (15) and (16), the ↗in incredulous RDs conveys S’s lack of commitment to the propositional content of the declarative, 𝑝. Moreover, since 388 12 Everything that rises must converge they are inquisitive RDs, they require contextual evidence for 𝑝. It is the contrast between the prior not-𝑝bias and this contextual evidence for 𝑝that makes the speakers of incredulous RDs incredulous. (21) exhibits one more kind of incredulous RD that features a prior not-𝑝speaker bias like (8) and (20), but that differs from them in that the contextual evidence immediately settles the issue in favor of 𝑝, which creates a challenge for applying the non-commitment analysis of ↗in (15) to it.19 (21) S believes that her friend A is abroad on vacation and not due back for some time. Then, S bumps into A at the local café. S: You’re back↗ The truth of 𝑝in (21) is so obvious to S that she can’t help but be publicly committed to it. Alexander Williams noted to me (p.c.) that her use of the pronoun you even shows that she has updated her discourse model with A’s presence. How can S’s lack of commitment to 𝑝as required by ↗according to (15) and (16) be satisfied in (21)? One possible response is to say that it can’t, similar to the case of assertive RDs like (10), and so lack of commitment is interpreted to be about some other relevant proposition 𝑞. However, this would be counterintuitive, since it would force us to give distinct analyses for different kinds of incredulous RDs: those like (21) would be assertions while those like (20) would be questions.20 This is undesirable since (8), (20), and (21) are intuitively all very similar to one another. All three could be paraphrased as “S is incredulous that 𝑝”, so we’d like to have a unified analysis of them. The solution is to extend the analysis I gave for (8) and (20) to (21) by taking (21) to involve pretense: while S clearly knows that 𝑝is true, she is nevertheless shocked at 𝑝because just prior to new evidence supporting 𝑝, S would have happily committed publicly to not-𝑝. Thus S conveys her surprise via the lack of commitment expressed by ↗. Compare this to exclamations of I can’t believe it! or I can’t believe you’re here! when the proposition is evidently true (thanks to an anonymous reviewer and to Alexander Williams for this comparison). 4.3 A similar illocutionary ambiguity in rising imperatives Rudin (2018) observes that ↗can appear with imperatives, and that it has a roughly similar effect as with declaratives and interrogatives, conveying a lack 19Thanks to an anonymous reviewer for Sinn und Bedeutung for this example. 20Treating all incredulous RDs as assertions, on the other hand, appears even less likely to work, since examples like (20) could never be analyzed as S asserting 𝑝. 389 Daniel Goodhue of commitment that manifests as suggested actions that the addressee could take but is not required to. For example, consider the imperatives used by the boss in (22a), which have standard list intonation on the first two (either a slight rise to a plateau or a rise-fall-rise), culminating in a final fall (H* L-L%, ‘↘’), and compare them to the rising imperatives used by the coworker in (22b): (22) New employee: What should I do now? a. Boss: Take the trash out, wash the sink, (then) take your break↘ b. Coworker: Take the trash out↗wash the sink↗take your break↗ The boss is issuing a set of commands to be carried out in a particular order, while the coworker is merely offering a menu of suggested options. Now consider the boss’s use of a rising imperative in (23): (23) The boss has just told the new employee to replace the trash bags in all of the trash cans, tie up the used ones, and put them by the back door. After doing this, the employee asks: New employee: What should I do now? Boss: Put them in the dumpster↗ The boss’s rising imperative in (23) could be interpreted as a weak suggestion if we assume that the boss is either inappropriately negligent or inappropriately unauthoritative. However, another possible interpretation is that the boss is issuing a normal command, and that ↗raises another issue, roughly How do you not know this?, or more rudely Are you stupid? (indeed, the use of ↗in (23) has the effect of insulting and belittling the employee). An explanation parallel to that given for the inquisitive/assertive RD split above can be given here, namely the boss has the social authority to issue commands while the coworker does not. This explains why the most natural interpretation of the rising imperatives in (22b) is a weakening of the commands themselves, while one natural interpretation of the rising imperative in (23) is that it retains its usual commanding force, and ↗raises another issue that the audience is left to infer pragmatically. I leave a more formal exploration of the interaction of my account of ↗with imperatives to future work. But we can already see the family resemblance to the analysis of declaratives: In both declaratives and imperatives, ↗’s meaning usually targets the content of the clause it appears with, but it doesn’t have to. Whether the audience is cued to search for some other relevant content for ↗ to interact with depends on whether the context renders its application to the clausal content improbable. 390 12 Everything that rises must converge 5 Deriving assertion In asserting 𝑝, whether via a falling or rising declarative, the speaker S intends to commit to 𝑝, and so from S’s perspective, commitment is achieved from the moment the utterance is complete. Likewise, from the perspective of the addressee A, as soon as A has understood S’s utterance as an assertion, A will take S to be committed to 𝑝starting with whatever context immediately follows the utterance. My account so far leaves open how this assertive commitment comes about. Dynamic semantics/pragmatics models this by treating utterances as functions from contexts to contexts, and assertions in particular as updating discourse commitments in the output context 𝑐𝑛+1to reflect the new assertive commitment. The utterance function as I defined it in (14) is at odds with this view, since it only adds clausal content to the table and has no effect on commitments. This was done intentionally to avoid building commitment into a conventional discourse effect for particular clause types or intonations (pace Farkas & Bruce 2010, Lauer 2013, Farkas & Roelofsen 2017, Jeong 2018, Rudin 2018, 2022). Instead, my aim was to separate out linguistic meaning from context update, creating space for pragmatics to operate on linguistic meaning and context to produce assertive force. On this view, (14) can be thought of as a special addition to locutionary force. When S utters 𝜙, the context is updated with the fact of S having said it. This update includes the syntactic structure and the compositional interpretation computed on it, including the contribution of the prosodic contour, though not yet the full pragmatic consequences of any of these. Going beyond the usual locutionary act, I also take this update to include the addition of the content of 𝜙to the table 𝑇, changing that component of our formal model of context. It’s from this position in the conversation that the audience can, if need be, draw inferences to recover S’s intended illocutionary force. To see how this works for assertive force, consider S’s assertion of a falling declarative in (24): (24) A: How old is your daughter? S: She’s nine↘ Recall that ‘↘’ stands for the the falling intonation typical of assertions of declaratives, H* L-L%. One approach to the analysis of assertive commitment in falling declaratives would be to treat ↘as conveying commitment, an opposite counterpart of ↗. However, I reject this view and instead treat ↘as a meaningless default. There are a few reasons for this. First, ↘does not impose a sufficient condition for the presence of assertive commitment, as it is also the 391 Daniel Goodhue standard contour for constituent questions.21 Beyond this, ↘also appears in imperative commands/requests, as noted in Section 4.3, and below I will discuss its presence in falling declarative questions (Bartels 1999, Gunlogson 2008). Given the breadth of ↘’s distribution, it’s hard to see how to maintain a unified commitment analysis. ↘appears to be a default elsewhere intonation. Second, ↘ does not impose a necessary condition on the presence of assertive commitment, as other, non-falling contours can be used in assertive utterances. For example the rise-fall-rise contour (Ward & Hirschberg 1985) and the contradiction contour (Liberman & Sag 1974) are two well-studied rising contours that are used in assertions that commit the speaker to the declarative content. Put otherwise, analyzing falling and rising intonation as manipulating the presence or absence of speaker commitment produces an empirically inadequate dichotomy.22 To maintain an intonation-encodes-commitment view while accounting for these other contours, a list of contours that speakers use to make commitments would need to be made. But if commitment is the elsewhere case, it suggests that it shouldn’t be baked into lexical meanings for contours. Third, the assertive RDs discussed in this paper are another case of a non-falling contour appearing in assertive utterances. But in this case, the contour, ↗, cannot be added to a list of commitment contours since it frequently appears in inquisitive speech acts. For these reasons, I don’t want to lean on ↘to explain assertive commitment. Even if it did play a direct role in producing commitment in all utterances it appears in (which it doesn’t), it would still leave assertive commitment unexplained in other utterances that ↘does not appear in. Instead I will pursue an approach in which assertive commitment is derived. An anonymous reviewer asked why I don’t do this the other way around, arguing that it is more intuitive to have ↘convey commitment and derive lack of commitment from ↘’s absence (the reviewer is not the only linguist to have made this argument to me). While the preceding paragraph gives several sufficient reasons not to imbue ↘with commitment, my approach can also be defended on the grounds of the Stalnakerian view of conversation I have assumed. Conversation is about increasing understanding of how the world is, and commitment plays a central role in achieving this goal. So commitment is part of the normal course of conversational events, and is in that sense relatively unmarked. Lack of 21Bartels (1999) and Truckenbrodt (2009) make the argument that constituent questions give rise to an existential presupposition, which ↘expresses commitment to. Whatever the merits of this argument are, there are other reasons below to question that ↘conveys commitment. 22Cf. Rudin (2018) for an example of an analysis along these lines. Rudin (2022) revises the analysis so that commitment is built into all utterances (within a restricted domain), and ↗ cancels that commitment. 392 12 Everything that rises must converge commitment, on the other hand, presents a detour on the road to achieving the conversational goal. Sometimes detours are necessary, but they are best avoided if possible. Thus the choice to convey lack of commitment is relatively less likely, and therefore more marked, and so more likely to be marked with a meaningful contour than commitment is. So how is assertive commitment to the proposition expressed in (24) that she is nine, call it 𝑝, derived? In a nutshell, if S had intended to convey lack of commitment to 𝑝, S should have marked that with ↗; since S didn’t do so, S must intend to commit to 𝑝. In greater detail: According to (14), S adds the singleton set of 𝑝to the table 𝑇. There is nothing in the meaning of the declarative clause itself, or in ↘, or in the utterance function that implies S’s commitment to 𝑝. Given the broadly Stalnakerian view of conversation adopted here, and the support requirement in (17) in particular, I assume that by adding {𝑝} to 𝑇, S either intends to commit to 𝑝or not (i.e., S intends a lack of commitment to 𝑝). I also assume a requirement to maximize non-at-issue content (MaxNAI), comparable to Gricean Quantity and Maximize Presupposition, that says that if S intends to convey certain non-at-issue content 𝛾, and there is a linguistic expression 𝑙that conveys 𝛾, then an utterance 𝜙that includes 𝑙is preferred to an alternative utterance 𝜓that does not include 𝑙but is otherwise identical. MaxNAI can render ↗preferred to ↘in appropriate utterances: There is a linguistic form, ↗, that explicitly conveys lack of commitment. So if S intends to convey lack of commitment to 𝑝while uttering declarative 𝜙, S should explicitly do so via ↗. In the case of the falling declarative in (24), S has chosen not to use ↗, and so we can infer that S doesn’t intend lack of commitment to 𝑝. But since, in uttering 𝜙, S either intends commitment to 𝑝or lack of commitment to 𝑝, it follows that S intends to commit to 𝑝. Here is the same train of reasoning, in schematic form: 1. S utters “She’s nine↘” (24) •¬S explicitly conveys 𝑝∉DCS(consequence of (1)) 2. 𝑇+ {𝑝} ((1), utterance function in (14)) 3. S intends 𝑝∈DCS∨S intends 𝑝∉DCS((2), Stalnakerian pragmatics/(17)) 4. S′utters “She’s nine↗” (NAI-stronger alternative to (1)) • S′explicitly conveys 𝑝∉DCS′(consequence of (4)) 5. S intends 𝑝∉DCS→S explicitly conveys 𝑝∉DCS((1), (4), MaxNAI) 393 Daniel Goodhue 6. ¬S intends 𝑝∉DCS(modus tollens on (1) & (5)) 7. S intends 𝑝∈DCS(disjunctive syllogism on (3) & (6)) This is how assertive commitment can be derived for falling declaratives based on the proposed meaning of ↗in (15), combined with a broadly Stalnakerian view of conversation, including the support requirement in (17), as well as Gricean pragmatics. Given that assertive commitment is essentially an implicature on this view, it’s reasonable to wonder if it can be canceled. The answer may be yes, and the evidence comes from falling declarative questions: (25) A and S are colloquium committee organizers. A is in charge of today’s colloquium, so both A and S know that A has more information about what is going on with it than S. A: We have a problem: We need someone to go pick the invited speaker up from the airport, but Kate is on the other side of town. a. S: And James isn’t available↘ b. S: And James isn’t available↗ From A’s claim that they have a problem, S is able to infer that the people who would usually be asked to pick the speaker up from the airport are unavailable. But A only mentioned Kate, so S asks about the other usual person James. The felicity of the falling version in (25a) shows that falling declaratives are not always assertions, which suggests a cancellation of the derivation sketched above (cf. Bartels 1999: 243 and Gunlogson 2008 for more discussion of falling declarative questions). However, the felicity of the rising version in (25b) raises a question for the account I gave above: Why doesn’t MaxNAI make the use of ↗necessary for an inquisitive interpretation? Note that there is an intuitive difference between (25a) and (25b): In (25a), S seems to be pretty confident about the conclusion they have drawn that James isn’t available based on A’s claim that they have a problem; but given A’s epistemic authority, S wants A to confirm this conclusion, and so S’s utterance is still a question looking for a yes or no response. In (25b), on the other hand, S seems a little less confident that this inference about James is true. So the choice to use ↘instead of ↗in (25a) still has an interpretational effect, even if it doesn’t lead to full assertive commitment. The reason (24) leads to full assertion while (25a) does not is that A is the clear epistemic authority 394 12 Everything that rises must converge in (25); A’s authority disrupts the derivation of assertion (25).23 In Gunlogson’s (2008) terms, S is making a dependent commitment in (25a). While these remarks remain preliminary, the existence of falling declarative questions, distinct from both falling declarative assertions and rising declaratives, further suggests that it is correct to derive assertion pragmatically instead of baking it into the meaning of the declarative+↘pairing. 6 Conclusion The advantage of the account I have proposed is its ability to explain how we arrive at distinct illocutionary forces when interpreting one and the same linguistic form. I have posited unitary meanings for linguistic forms (declaratives denote 𝑝, polar interrogatives denote {𝑝,¬𝑝},↗conveys lack of commitment to a relevant proposition 𝑞), as well as a single utterance function that adds utterance content to the table. Speakers can employ the combination of declarative and ↗to assert or to question, thanks primarily to the ability of ↗to apply directly to the declarative content or not. I further observed that pragmatic pressure to grow the common ground in turn leads to pragmatic pressure to raise and resolve issues via interlocutor support for a proposition in the issue. This combined with the meanings of the linguistic forms enables speakers to implicate, and addressees to derive, the distinct discourse effects of inquisitive and assertive RDs such as She’s nine↗in (9) and (10), as well as falling declaratives as in (24). Illocutionary force on this view resides purely in the pragmatics, with the audience’s ability to recover the speaker’s intended force depending only in part on input from the linguistic system. 6.1 Comparison to prior accounts Farkas & Roelofsen (2017), Jeong (2018), and Rudin (2018, 2022) produce distinct accounts that nevertheless arrive at the same conclusion that clause type +intonation determines illocutionary force as a matter of convention. Farkas & Roelofsen (2017) argue that ↗is a semantic operator that forms polar interrogatives. 23Relative authority then seems to play a central role in the force interpretation of utterances, especially if the intonational form of the utterance would otherwise usually be used to convey the opposite force: Assertive rising declaratives are identified in cases where S is the authority on 𝑝, while falling declarative questions are identified in cases where A is the authority on 𝑝. Cf. a similar observation in Jeong (2018), and see Mosegaard Hansen (2001) for a related observation about the use of French polar interrogatives with declarative word order from a conversation analytic perspective. 395