What can Cantonese sentence-final particles tell us about rhetorical questions?

Chapter 7 What can Cantonese sentence-final particles tell us about rhetorical questions? Angelika Kiss University of Toronto Roger Yu-Hsiang Lo University of British Columbia Justin R. Leung University of Toronto Rhetorical questions have been analyzed both as utterances equivalent to assertions and as questions. We acknowledge that they have the characteristics of both: their hybrid nature is best captured if we give them an inquisitive semantic treatment. Rhetorical questions can suggest both an empty set answer and a non-empty set answer, and this fact gives rise to an asymmetry in terms of their inquisitiveness and informativity, which is also reflected in their prosody. In a perception experiment, Cantonese speakers associated information-seeking questions and rhetorical questions suggesting an empty set answer with a certain intonational contour on the sentence-final particle, but there was no clear pattern for rhetorical questions with a non-empty set answer. We explain this pattern by the informativity/inquisitiveness of these utterances, which maps to different levels of dependence from the common ground. 1 Introduction Both the meaning and the form of rhetorical questions have received much attention in the literature. As for their meaning, they were first considered in general Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung. 2025. What can Cantonese sentencefinal particles tell us about rhetorical questions? In Tue Trinh, Anton Benz, Daniel Goodhue, Kazuko Yatsushiro & Manfred Krifka (eds.), Biased questions: Experimental results and theoretical modelling, 207–240. Berlin: Language Science Press. DOI: 10.5281/zenodo.17158188 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung to have “the illocutionary force of an assertion of the opposite polarity from what is apparently asked” (Han 2002: 201). In this sense, Han’s example in (1), when asked as a rhetorical question, means that everybody likes ice-cream. Since under this reading, the set of people who do not like ice-cream is empty, we refer to such rhetorical questions as empty set rhetorical questions. (1) After all, who doesn’t like ice-cream? (Caponigro & Sprouse 2007: 12) Later, it has become clear that not all rhetorical questions denote such an empty set answer, a fact also acknowledged by Han (2002: fn 6), because they can denote a non-empty set answer as well. Under such a reading, (1) would be used in a context where it is common ground that someone, known to both the speaker and the addressee, does not like ice-cream. In subsequent studies, rhetorical questions are defined as questions with an obvious answer (Rohde 2006, Caponigro & Sprouse 2007, Biezma & Rawlins 2017). In these accounts, the obviousness of the answer plays a major role, but not the kind of that obvious answer (i.e., empty set or non-empty set answer). Rhetorical questions therefore would form a homogeneous group. Given the homogeneity in terms of meaning, we expect that rhetorical questions will also form a homogeneous group in terms of their prosody. This is implicitly assumed by Biezma & Rawlins (2017), who refer to a “characteristic (hard to pin down) rhetorical prosody” that characterizes rhetorical questions regardless of the kind of answer they expect. A number of recent production studies have shown that the prosody of empty set rhetorical questions indeed distinguishes them from information-seeking questions (see Braun et al. 2018 for German, Dehé & Braun 2019 for English, Dehé & Braun 2020 for Icelandic, and Zahner et al. 2020 for Mandarin). But since these studies do not consider the class of non-empty set rhetorical questions, they neither support nor challenge the assumption that the two types of rhetorical questions have the same prosodic form. However, based on even more recent production experiments by Lo et al. (2019) on Cantonese and Lo & Kiss (2020) on Mandarin, we have reason to believe that the kind of answer rhetorical questions suggest plays a role in their semantics. When participants read the same interrogative in three different contexts – one that gives the interrogative an information-seeking question reading, one that gives it an empty set rhetorical question reading, and one that gives it a nonempty set rhetorical reading – the utterances show a three-way distinction in certain prosodic properties, a contrast that is not predicted by any of the abovementioned accounts. 208 7 Cantonese sentence-final particles & rhetorical questions If the two kinds of rhetorical questions are marked differently in their form, both from each other and from information-seeking questions, it suggests that the two types of rhetorical questions differ in their semantics. This is the conclusion of Kiss & Lo (2021), who offer an inquisitive semantic analysis that describes the three question types as expressing three distinct kinds of speaker commitment because they form a scale in terms of how informative/inquisitive they are. Namely, information-seeking questions are the least informative, and empty set rhetorical questions are the most informative among the questions on their scale, and non-empty set rhetorical questions are between them. In this chapter, we test these claims by means of a perception experiment done with native speakers of Cantonese; participants hear string-identical interrogatives which all have their sentence-final particle aa3prosodically manipulated.1 The experiment allows us to test whether there is indeed a three-way prosodic distinction among these question types, and indirectly, it allows us to see if the three question types indeed form a scale in terms of their informativity / inquisitiveness, which maps onto different degrees of context-dependence, which in turn may affect how strongly speakers associate a certain contour with one or the other question type. To our knowledge, ours is the first study to compare both empty set and nonempty set rhetorical questions to each other and to information-seeking questions in a perception experiment. As such, it contributes to what we know about rhetorical questions, and to what we know about intonation and the aa-family of sentence-final particles in Cantonese. The paper is organized as follows. In Section 2, we present arguments for the claim that rhetorical questions are questions, and distinguish empty set and nonempty set rhetorical questions. In Section 3, we present Farkas & Roelofsen’s (2017) inquisitive semantic analysis of questions and extend it following Kiss & Lo (2021) so that it comprises information-seeking and rhetorical wh-questions. Section 4 presents previous work on the prosody of rhetorical questions, including Cantonese ones. In Section 5, the perception experiment is described, which is followed by the general discussion in Section 6 and by the conclusion in Section 7. 1The numbers 1 and 3 indicate the lexical tone of the particles. 1 is high level, 2 is high rising, 3 is mid level, 4 is low falling, 5 is low rising, and 6 is low level. See Matthews & Yip (2011) for a detailed characterization of Cantonese tones. 209 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung 2 Rhetorical questions In Section 2.1, we argue that rhetorical questions are better treated as (biased) questions, as opposed to assertions, and in Section 2.2, we show subtypes of rhetorical questions which differ in their meaning. 2.1 Rhetorical questions as questions It is widely known that rhetorical questions behave differently from genuine or information-seeking questions in discourse, namely they exhibit quite a few properties typically associated with assertions. Using the tests of Sadock (1971, 1974), who differentiates what he calls “queclaratives” from information-seeking questions, Han (2002) showed that rhetorical questions, but not informationseeking ones, can be introduced by after all and can be followed by an utterance that is introduced by yet. (2) a. After all, do phonemes have anything to do with language? b. Do phonemes have anything to do with language? Yet people continue to believe in them. Sadock (1971: 225), cited by Han (2002) After all and yet can felicitously occur in assertions, but not in informationseeking questions; yet when the interrogative Do phonemes have anything to do with language? is interpreted as a rhetorical question, it can appear in such environments. Another assertion-like property rhetorical questions exhibit is their compatibility with strong negative polarity items or minimizers (Han 2002, Abels 2003, Guerzoni 2004, Biezma & Rawlins 2017). (3) When did John lift a finger to help us? Minimizers are felicitous in assertions but not in information-seeking questions. The interrogative in (3) can only have a rhetorical question interpretation, conveying that ‘John never helped us’; thus rhetorical questions behave like assertions in this respect as well. These and further arguments gave rise to the view that rhetorical questions are assertions (Han 2002), a view that has been challenged by the observations of Rohde (2006) and Caponigro & Sprouse (2007). The latter authors pointed out some inherently question-like properties of rhetorical questions that receive no explanation if they are treated as assertions. Most importantly, rhetorical questions can also suggest a non-empty answer as shown in (4), not necessarily an empty set answer as in examples (2) and (3). 210 7 Cantonese sentence-final particles & rhetorical questions (4) Who has fed you and given you a proper education? (Context: a mother to her son) Rohde (2006) citing Han (1998) In addition, rhetorical questions can be answered the same way as questions, regardless of the kind of answer they suggest, as shown by the following minimal pair in (5). (5) a. Speaker: You should stop saying that Luca didn’t like the party last night. After all, who was the only one that was still dancing at 3am? Addressee or Speaker: Luca. b. Speaker: You should stop saying that Luca didn’t like the party last night. After all, Luca was the only one that was still dancing at 3am! Addressee or Speaker: #Luca. (Caponigro & Sprouse 2007: 124) If rhetorical questions function as assertions, they should be felicitous as answers to information-seeking questions, as in (6a), but this is not always the case, as Biezma & Rawlins (2017) point out. (6) a. A: Does Ed McMahon drink? B: Is the Pope a Catholic? (Schaffer 2005: 438) b. A: Does Mother Teresa drink? B: #Does Mother Teresa drink? (as a rhetorical question) While A receives an affirmative answer from B in (6a) who uses the rhetorical reading of Is the Pope a Catholic?, (6b) shows that if both the information-seeking question and the rhetorical question given in response are conveyed by the same interrogative, the result is odd. B’s utterance in (6b) can at most be interpreted as an echo question, but not a rhetorical question conveying ‘Mother Teresa does not drink’, which, under the assertion-like analysis, should be an acceptable answer. However, if we consider the proposal of those who treat rhetorical questions as questions with an obvious answer (Rohde 2006, Caponigro & Sprouse 2007, Biezma & Rawlins 2017), the infelicity of (6b)-B receives a natural explanation. Speaker A raised the issue of whether Mother Teresa drinks, after which B’s response is infelicitous for two reasons. First, because B’s utterance is a question, too, so despite the obvious answer, it still raises the same issue, which is an infelicitous move. Second, since A’s utterance is an information-seeking question, it is evidence that the issue of ‘whether Mother Teresa drinks’ is not settled. If so, the answer to B’s utterance should not be treated as common ground. 211 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung Treating rhetorical questions as questions with an obvious answer has further advantages. Unlike in the assertion-like analysis, both rhetorical questions suggesting an empty set answer and ones with a non-empty set answer are included; and it explains why rhetorical questions can be answered. Lastly, Caponigro & Sprouse (2007) point out that rhetorical questions are also like informationseeking questions in that they can have multiple wh-phrases and they can also be embedded. However, the question-like analysis cannot explain why minimizers can occur in rhetorical questions if the suggested answer is the empty set, to which the assertion-like analysis offers a solution. We nevertheless adopt a question-like analysis of rhetorical questions, and propose that to account for their assertionlike properties, we need to take into account the kind of answer suggested by the utterance (i.e., whether it is an empty set answer or a non-empty answer). 2.2 Rhetorical questions are not all alike Rhetorical questions thus can convey an empty set answer or a non-empty set answer, as shown in examples (7b) and (8b), respectively; however, this distinction does not play a role in the analyses that treat them as questions (Rohde 2006, Caponigro & Sprouse 2007, Biezma & Rawlins 2017). (7) Context: David is considering helping his colleagues. a. David: Should I help them? b. Erez: No way! Who helped you when you were in trouble? (8) Context: It is common ground that Cleo helped David when he was in trouble. a. David: Who should I trust? b. Erez: Well, who helped you when you were in trouble? (Of course Cleo!) The only exception known to us is the account of Jamieson (2018), who treats the two rhetorical question types differently. Jamieson (2018) refers to empty set rhetorical questions as generic rhetorical questions. This generic interpretation is due to a metavariable 𝜖, which affects the semantic value of the wh-phrase so that it no longer denotes the ordinary type of domain it is associated with, that is, who does not denote humans, where does not denote locations, etc. (9). The metavariable guarantees that the answer set becomes empty under its ordinary value, and so the rhetorical question is interpreted as an assertion, as in Han’s 212 7 Cantonese sentence-final particles & rhetorical questions approach. Thus, Erez’s utterance in (7b) amounts to saying ‘Nobody helped you when you were in trouble’. (9) a. JwhoK𝑜=𝜆𝑤.𝜖∉human in 𝑤 b. JwhereK𝑜=𝜆𝑤.𝜖∉location in 𝑤(Jamieson 2018: 324) Rhetorical questions that suggest a non-empty answer, on the other hand, are referred to by Jamieson as pragmatic rhetorical questions: these are questions which owe their assertion-like properties to the context, as they are questions with an already known answer. In agreement with Jamieson (2018), we posit that the two types of rhetorical questions have distinct semantics. 3 Rhetorical questions in inquisitive semantics Previous analyses of rhetorical questions have made it clear that this question type shares several characteristics with assertions, as well as with informationseeking questions. Rhetorical questions are not the only question type with both assertion-like and question-like properties. Tag questions, consisting of a declarative anchor and a reduced interrogative clause, have been analyzed as a hybrid speech act type, being an assertion and a question at the same time, and declarative questions (also known as rising declaratives), which consist of a declarative clause and bear a question-like intonation, share properties of both as well (Gunlogson 2003, Asher & Reese 2007, Farkas & Roelofsen 2017). The way Farkas & Roelofsen (2017) analyze such non-canonical questions is particularly important for our purposes because the inquisitive semantic framework they use makes it possible to capture both their assertion-like and questionlike nature. In addition, they make falsifiable claims about the contribution of the marked form to utterance meaning. We are primarily interested in wh-interrogatives. This is because, polar interrogatives suggesting a positive answer, which would correspond to rhetorical questions with a non-empty set answer, proved to be considerably more problematic to elicit in experiemntal settings. That is, in experimental settings, these were much harder to elicit than ones that suggest the empty set as an answer. Polar rhetorical questions suggesting a positive answer nevertheless do exist, as Rohde (2006) reports, a corpus example of which is shown in (10). (10) Has the educational system been so watered down that anybody who’s above average is now gifted? (Rohde 2006: 135) 213 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung While Farkas & Roelofsen (2017) only treat utterance types with a sentence radical, they claim that wh-interrogatives, too, can fit in their model. In this paper, we adopt the proposal of Kiss & Lo (2021) (see Section 3.2) who offer an extension of their inquisitive semantic framework to accommodate wh-interrogatives conveying information-seeking and rhetorical questions, but our claims arguably also hold for polar interrogatives. 3.1 Inquisitive semantics 3.1.1 Basic notions In Farkas & Roelofsen (2017), a proposition 𝑃is modeled not only as a set of possible worlds (Stalnaker 1978) but also as a set of information states, which themselves are sets of possible worlds. Besides their inquisitive content, propositions are also characterized by their informative content, marked as info(𝑃), which is a set of those possible worlds that are members of any of the information states within the given proposition. The inquisitive content of a proposition expresses the issue it raises, and the informative content expresses which possible worlds, if any, can be discarded from the speaker’s commitment set. Any information state 𝑠can have a substate 𝑡such that if a possible world is a member of 𝑡, it is also a member of 𝑠. Substates within an information state form a lattice, and their maximal element is referred to as an alternative. Whenever the possible worlds that make up a proposition 𝑃’s informative content do not form a single alternative, which would be a member of 𝑃’s inquisitive content, 𝑃 is inquisitive. And whenever the informative content of 𝑃is a proper subset of the set of all possible worlds 𝑊,𝑃is informative. (11) Informative and inquisitive propositions a. A proposition 𝑃is inquisitive iff info(𝑃)∉𝑃 b. A proposition 𝑃is informative iff info(𝑃)≠𝑊(Ciardelli et al. 2019: 23) A declarative sentence conveying an assertion consists of one alternative and is therefore informative and non-inquisitive, whereas a polar interrogative conveying an information-seeking question consists of two alternatives, and is therefore inquisitive and non-informative. Both assertions and polar questions have sentence radicals, which we assume, following Stenius (1967), is the part of a declarative or interrogative sentence that signifies its “descriptive content”. Utterances with a sentence radical have what Farkas & Roelofsen (2017) call a highlighted alternative, marked by 𝛼, which is the alternative denoted by the sentence radical. 214 7 Cantonese sentence-final particles & rhetorical questions (12) Inquisitive and informative content of a declarative conveying an assertion a. 𝑃= {𝛼}↓ b. info(𝑃) = info(𝛼) (13) Inquisitive and informative content of a polar interrogative conveying an information-seeking question a. 𝑃= {𝛼,𝛼}↓ b. info(𝑃) = 𝑊 The basic discourse context proposed by Farkas & Roelofsen keeps track of the participants, the table, and the commitment sets assigned to each participant, as outlined in (14). (14) A basic discourse context is a triple ⟨participants, table, commitments⟩, where: a. Participants: the set of discourse participants; b. Table: a stack of propositions, representing the proposals made so far; c. Commitments: a function that maps every participant 𝑥∈Participants to a set of possibilities, those possibilities that 𝑥is publicly committed to. (Farkas & Roelofsen 2017: 255) “The common ground” is the locus of mutual commitments (Stalnaker 1978, Farkas & Bruce 2010), which here is derived from the participants’ commitment sets. (15) a. Commitment set: 𝑐𝑠𝑥=⋂Commitments(𝑥) b. Common ground: 𝑐𝑔=⋃{𝑐𝑠(𝑥)|𝑥∈Participants} (Farkas & Roelofsen 2017: 255) 3.1.2 Semantic interpretation The elegance of Farkas & Roelofsen’s (2017) account lies in differentiating what they refer to as the basic conventional discourse effects from the special discourse effects of an utterance, which is summarized by the division of labor principle. (16) Division of labor principle a. The discourse effects of unmarked forms should be fully determined by their semantic content and the basic convention of use, 𝐹𝑏. 215 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung The informative content of empty set rhetorical questions are comparable to the informative content of assertions, but they are still inquisitive because of their interrogative clause. Non-empty set rhetorical questions are less informative than empty set rhetorical questions, because even though they discard Α, they still contain more than one alternative. Even though informativity is a binary property, it can be treated as a gradient one in the sense that the more alternatives an expression contains, the less informative it is. Similarly, even though inquisitiveness is a binary property of expressions, the scale in Table 1 may be seen as a scale of some sort of global inquisitiveness. Assertions are not inquisitive in any sense; empty set rhetorical questions are only inquisitive by virtue of their syntactic form; non-empty set rhetorical questions are inquisitive both by virtue of their syntactic form and of their meaning, because they denote more than one alternative yet those do not cover logical space; and finally, informationseeking questions are inquisitive and not informative, and as such, they represent the purest case of an inquisitive utterance on this scale. Note that this analysis only makes reference to the semantic properties of rhetorical questions, leaving out their pragmatic aspect, namely, the role of the common ground in their interpretation. Of course, rhetorical questions with a non-empty set answer become just as informative as rhetorical questions with an empty set answer, once the addressee has consulted the common ground. However, that empty set rhetorical questions are informative at a different level, as they are informative even without knowing the common ground. We argue that it is this kind of meaning that gives rise to markedness. We assume that if a certain meaning trait gets marked in an utterance type, it has to be one that is generalizable across contexts. The meaning of informationseeking questions is generalizable across contexts as it denotes, in any context, the maximal number of alternatives. Similarly, the meaning of empty set rhetorical questions is generalizable across contexts, because any domain has a complement. The meaning of non-empty set rhetorical questions, however, is a certain answer choice from a set: which answer it is depends entirely on the given common ground, it is therefore not generalizable across contexts in the same sense. A meaning trait that is generalizable is less dependent on the particular context of utterance, and as such, it is more likely to be marked in a conventionalized way. Prosody is one way by which utterance types with differences in speaker meaning can be distinguished. Assuming that there is a systematic relationship between markedness in form and special discourse effects, as proposed by Farkas & Roelofsen (2017), we expect to find prosodic correlates of these semantic differences. In what follows, we move on from discussing the meaning of rhetorical questions to discussing their form. 222 7 Cantonese sentence-final particles & rhetorical questions 4 Rhetorical questions in Cantonese Prosody is one of the ways in which the speaker’s epistemic stance is conveyed. Speaker meaning is especially important in biased questions; therefore, biased questions are often prosodically marked. Given that the same interrogative can be used both as an information-seeking or as a rhetorical question, it is expected that the same interrogative will assign different prosodic marking to the different question types it conveys. 4.1 Previous work on the prosody of rhetorical questions Most studies compare information-seeking questions to rhetorical questions with an empty set answer. They unanimously point to the conclusion that all else being equal, the two question types have distinct prosody, although the way they differ is specific for each language. There are two issues that we point out in relation to this body of work. First, prosodic studies on rhetorical questions typically do not establish formmeaning relations, though some of them do refer back to the introspective judgments of Han (2002), who does propose such relations. Han (2002: 215) claims that a rhetorical question (by which she means a rhetorical question with an empty set reading) has “an intonational contour of an assertion”, unlike an information-seeking question, which has “a rising intonation”. While it is not clear whether Han intended her claims to be interpreted for English or for any language, her intuitions receive only partial support by recent empirical studies on various languages. Rhetorical questions indeed tend to have a low final boundary tone (L%) compared to information-seeking questions. This tendency has been observed in English polar interrogatives: a corpus study on English polar questions shows that while both information-seeking and rhetorical questions can end with L%, there is more variation between H% and L% in information-seeking questions (Banuazizi & Creswell 1999); and the experimental results of Dehé & Braun (2019) also confirm that the boundary tone H% in English is reserved for information-seeking polar questions, while L% is produced with an (empty set) rhetorical reading. German rhetorical wh-questions tend to end in L%, while there is more variation in the boundary tone of informationseeking wh-questions (Braun et al. 2018). In Estonian, H% is found significantly more often in information-seeking questions than in (empty set) rhetorical questions (Asu et al. 2020). However, Dehé & Braun (2019) also observe that English wh-questions can end with L% regardless of which reading is produced; Dehé & Braun (2020) note that the boundary tone is always L% across utterance types in Icelandic; and Braun et 223 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung al. (2018) even claim that the boundary tone alone is not sufficient to determine whether an utterance is an information-seeking or a rhetorical question. In addition, most of these studies underline that prosodic cues other than pitch, such as voice quality and duration, too, participate in marking utterance type. For example, Braun et al. (2018) and Dehé & Braun (2019) have found that rhetorical questions (with an empty set reading) in German and English tend to be produced in a breathier voice compared to information-seeking questions, and that rhetorical questions overall tend to have a longer duration for certain constituents or the entire utterance (Braun et al. 2018, Dehé & Braun 2019, Asu et al. 2020). These observations show that the intonational differences between information-seeking and empty set rhetorical questions cannot be reduced to a difference between their sentence-final tunes. In sum, the only candidate for a form-meaning relation proposed in the literature, namely that information-seeking and empty set rhetorical questions differ from each other in their high versus low boundary tones, respectively, is challenged by empirical data. The second relevant issue that arises when theoretical work on rhetorical questions meets empirical work on this topic is that the majority of the prosodic literature does not consider non-empty set rhetorical questions. The first study known to us that compares both rhetorical question types to information-seeking questions is a production experiment on Cantonese rhetorical wh-questions (Lo et al. 2019). Unlike previous studies, Lo et al. (2019) elicited string-identical whinterrogatives in three different contexts so that one would favor an informationseeking question reading, another one, a non-empty set rhetorical question reading, and the third one, an empty set rhetorical question reading, see an example item in (28). (28) 有 jau5 have 邊個 bin1go3 who 想 soeng2 want 飲 jam2 drink 咖啡 gaa3fe1 coffee 呀 aa3? sfp ‘Who wants to drink coffee?’ a. Information-seeking question Context: You are having a family gathering in your home with a lot of people. After lunch, you want to serve tea and coffee, but you don’t know how much to prepare. You prepare the coffee first, and you want to find out how many cups are needed. So you ask: “I don’t know how many people want coffee. Who wants to drink coffee?” b. Empty set rhetorical question Context: Mary is throwing a party. It’s 1 am and everybody is about to 224 7 Cantonese sentence-final particles & rhetorical questions leave, but she is wondering if she should still make some coffee. You think nobody would drink any – some of them have already left anyway. So when Mary is telling you that she’ll start the coffee machine, you say: “Mary, wait, you shouldn’t start the machine. Who wants to drink coffee?” c. Non-empty set rhetorical question Context: Mary and you are roommates, and you had a guest from South America who brought you five different South American coffees. But neither you nor Mary drinks coffee, so you suggest to give all the coffee to someone. You just need to find a coffee drinker. The first person that comes in mind to both of you is John, your neighbor, because you both see him with his coffee on the balcony every morning, afternoon and evening – John is a real coffee-holic! So when Mary wonders who you should give all this coffee, you point to John, who’s drinking coffee on his balcony right now, and say: “Just look at that balcony. Who wants to drink coffee?” (Lo et al. 2019) Participants read and heard the context at the same time, after which they were given the reactive move in written form. Participants were supposed to read it out loud as if they were in a conversation with someone. The target question is always the last utterance within the move. Lo et al. (2019) found a three-way contrast in the prosodic realization of the sentence-final particles of the three question types. The production study of Lo & Kiss (2020) on the three question types in Northern Mandarin, which used the same method as Lo et al. (2019), also found a three-way contrast in their prosodic markedness. In sum, while there is a growing body of literature on the prosody of rhetorical questions, most are production studies, and most consider only the contrast between empty set rhetorical questions and information-seeking questions. As for Cantonese, the only production study involving both kinds of rhetorical questions is the one of Lo et al. (2019), whose results are shown in the next section. There is not one perception study known to us, in Cantonese or any other language, that compares both kinds of rhetorical questions to each other and to information-seeking questions. Thus the question whether the three question types – information-seeking questions, empty set and non-empty set rhetorical questions – can be distinguished in perception is left open. Answering this question is essential in testing our proposal on the semantics of rhetorical questions, because a perception experiment would show us whether there is a three-way distinction in prosody among the three question types, and in addition, it would show which question types are associated with a certain 225 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung prosodic form, which in turn is an indicator of whether they are generalizable across contexts or not. Before we turn to our perception experiment on Cantonese, we discuss intonation and the role of sentence-final particles in this language. 4.2 Intonation and sentence-final particles in Cantonese Cantonese has a large inventory of sentence-final particles, which appear at the end of utterances (and are therefore also known as utterance-final particles) and often convey affective and epistemic stance (Luke 1990, Matthews & Yip 2011, Sybesma & Li 2007, Wakefield 2019). Each sentence-final particle bears a lexical tone, which changes its meaning. Tonal languages make use of intonation as well; thus, the same channel – pitch – is used to convey both lexical meaning and utterance-level meaning. It has been claimed that sentence-final particles and intonation share the same function; that is, they divide labor in conveying speaker-oriented meaning. Wakefield (2012), for example, treats the function of the Cantonese sentence-final particle lo1as equivalent to the utterance-final rise in English declarative questions. Sharing function with intonation is not specific to sentence-final particles; it has also been observed for the wider class of discourse particles of which sentencefinal particles are a subtype (see Schubiger 1965 for German modal particles and Prieto & Roseano 2021 for Friulian and Catalan discourse particles). This division of labor between discourse particles and intonation and the considerable inventory of sentence-final particles in Cantonese together account for the fact that intonation is very restricted in this language compared to intonational languages. Due to this restrictedness, intonation tends to reveal itself on the last syllable of the utterance, which very often is the sentence-final particle itself. Interaction between tone and intonation on the last syllable of the utterance is therefore very strong; intonation can significantly alter the realization of its lexical tone, regardless of the contour of the tone and regardless of whether it is a lexical word or a sentence-final particle (Lee 2004, Ma et al. 2006a,b, Xu & Mok 2012, Tsui & Tong 2018, Wakefield 2019).5 5A reviewer notes that if the tone of the final syllable is overridden by intonation regardless of whether it is a sentence-final particle or a lexical word, it may be the case that it is prosodic form alone that matters, and a proper analysis can be built without making reference to sentencefinal particles. However, there is reason to believe that sentence-final particles (as well as other discourse particles in other languages) have the same function in conveying speaker meaning as intonation; the answer to this question is therefore not straightforward. Based on the results of Prieto & Roseano (2021), we may expect intonation to show more strongly in the absence of sentence-final particles. However, this needs to be tested empirically, and is for now left for future research. 226 7 Cantonese sentence-final particles & rhetorical questions From this, it is expected that when string-identical wh-interrogatives in Cantonese are uttered in different readings, they would differ from each other in the prosody of their last syllable. In fact, this was the case in the production experiment of Lo et al. (2019) who elicited information-seeking questions and the two types of rhetorical questions in Cantonese: the only significant differences in prosody (at least in terms of pitch and duration) were detected on the last syllable, which were the sentence-final particles aa3(呀) and aa1(吖). These two sentence-final particles, along with further members of the aafamily, are similar in function. According to Sybesma & Li (2007: 1760), aa3can occur in declaratives, interrogatives, imperatives and exclamatives, and it “makes the utterance fit more smoothly into the conversational context”. As for aa1, it is very much like aa3in terms of its function, but in addition, it “reveals more of the speaker’s emotion or attitude”. The following examples illustrate the difference. (29) a. cin4-min6 in front jau5 have hou2-do1 very-many jan4 people aa3 sfp ‘There are lots of people in front.’ b. cin4-min6 in front jau5 have hou2-do1 very-many jan4 people aa1 sfp ‘There are lots of people in front [why did you say there were just a few?].’ (Sybesma & Li 2007) (30) a. nei5 2s heoi3 go bin1-dou6 where aa3? sfp ‘Where are you going?’ b. nei5 2s heoi3 go bin1-dou6 where aa1? sfp ‘Where are you going?’ [challenging] (Sybesma & Li 2007) In Lo et al.’s (2019) experiment, both aa3and aa1were realized with low and level intonation when pronounced at the end of an information-seeking or empty set rhetorical question, and both were realized with rising pitch in non-empty set rhetorical questions. As for their duration, both aa1and aa3were significantly longer in empty set rhetorical questions than in other conditions. A summary of Lo et al.’s (2019) results is shown in Table 2. In sum, sentence-final particles in information-seeking questions were produced with a low, level pitch and a relatively short duration; in non-empty set rhetorical questions, with a rising pitch and a relatively short duration; and in 227 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung Table 2: The pitch and duration of the sentence-final particles aa1and aa3in Lo et al. (2019). Question type Pitch Duration Information-seeking questions (ISQ) low, level short Non-empty set rhetorical questions (RQ+) rising short Empty set rhetorical questions (RQ−) low, level long empty set rhetorical questions, with a low, level pitch and a relatively long duration. The two prosodic cues of pitch and duration were thus shown to play a role in production as they proved to be sufficient to distinguish the three question types prosodically. It remains to be seen, however, whether a three-way distinction is also present in perception, as the relevance of a prosodic cue is only as strong as the extent to which speakers rely on it in language processing. Thus a threeway distinction found in perception would corroborate those accounts that treat the meaning of information-seeking questions, empty set rhetorical questions and non-empty set rhetorical questions as different, that is, Jamieson (2018) and the one offered in this paper, as opposed to accounts that consider only a twoway distinction between information-seeking and rhetorical questions (Biezma & Rawlins 2017). And in case a three-way prosodic distinction is found, the experiment would also test the prediction that the less dependent an utterance type is on the common ground and the more generalizable across contexts, the more strongly the sentence-final particle is associated with a certain prosodic form. 5 Perception experiment We have proposed a semantic account that makes a distinction among two kinds of rhetorical questions (i.e., ones suggesting an empty set answer and ones suggesting a non-empty set answer) and information-seeking questions. This account has received support from production studies where a three-way distinction has been found (Lo et al. 2019, Lo & Kiss 2020). However, for a stronger support, we need to see if the three question types also differ from each other in perception. It has been shown by Kharaman et al. (2019) that German rhetorical questions are distinguished from informationseeking questions in perception, but the type of rhetorical question was not an independent variable (and it is not specified in the paper whether non-empty set 228 7 Cantonese sentence-final particles & rhetorical questions rhetorical questions were used at all). The question of whether the three question types are distinguishable from each other in perception still remains open. In what follows, we present our experiment in which we ask if speakers of Cantonese associate the three question types with distinct prosodic cues. If there is indeed a three-way distinction in meaning, as proposed by us and by Jamieson (2018), we expect a three-way distinction in form to be found in perception as well. On the other hand, if the two types of rhetorical questions are associated with the same prosodic cues, it corroborates the assumption made by the accounts in which rhetorical questions form a natural class based on the obviousness of the answer they suggest and nothing beyond that (Biezma & Rawlins 2017). In this case, rhetorical questions have the same meaning and, as a reviewer suggests, there could be some mechanism ensuring that in case a non-empty answer cannot be found in the common ground, the rhetorical question is by default interpreted as one with an empty set answer. If a three-way distinction is found in perception, however, there is a further prediction following from our account. Namely, speakers associate certain prosodic forms more readily with those question types that are interpreted more independently of the context, because those are the ones that are more likely to conventionalize their prosodic properties. As we have argued in Section 3.3, the interpretation of information-seeking questions and empty set rhetorical questions is less dependent on the particular context of utterance, because of their semantic properties: they are found on the edges of the informative/inquisitive scale for interrogatives (see the rightmost three columns of Table 1), which means that they are closest to being “purely” inquisitive and “purely” informative. Information-seeking questions express the maximum level of speaker ignorance: they are inquisitive and non-informative at the same time. Empty set rhetorical questions are inquisitive, too; however, their informative content is as informative as possible, since it is reduced to a single alternative. Thus the hearer does not need to consult the common ground in either of these cases: for informationseeking questions, all alternatives are elements of the meaning, and for empty set rhetorical questions, the informative content always covers the empty set answer alone, regardless of the domain. But the intended meaning of a non-empty set rhetorical question cannot be determined without making reference to the common ground. Note that this latter prediction, too, follows from the account of Jamieson (2018), who calls non-empty set rhetorical questions “pragmatic” rhetorical questions, for the same reason that it cannot be interpreted in the intended way unless the hearer recognizes the referent in the common ground as intended by the speaker. 229 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung Lastly, we expect participants to associate the three question types with prosodic forms with which they were produced in Lo et al. (2019) (see Table 2). However, all three question types looked at here can be marked both by aa1and aa3. While in production, we can control which sentence-final particle is presented to the participants, in our perception experiment the stimuli were presented auditorily only. Since we cannot know for sure which sentence-final particle they thought they heard, the expectation that we will find the same forms in perception that we found in production cannot be confidently predicted. 5.1 Stimuli Our stimuli were created from natural utterances produced by a female in her thirties, who is a native speaker of Hong Kong Cantonese and a linguist. The sentences were presented to her in written form, and she was instructed to produce them in a natural way as information-seeking questions. There were 12 sentences, all of which ended with the sentence-final particle aa3(呀). This particle was also used in Lo et al. (2019) for its compatibility with all three relevant question readings (Sybesma & Li 2007: 1760). An example is shown in (31). (31) 有 jau5 have 邊個 bin1go3 who 想 soeng2 want 飲 jam2 drink 咖啡 gaa3fe1 coffee 呀？ aa3? sfp ‘Who wants to drink coffee?’ The speaker read the information-seeking version of each utterance in a soundattenuated booth. The elicited utterances were then manipulated by changing the duration and pitch contour of the sentence-final particle. The duration and pitch dimensions were manipulated orthogonally, with the duration taking on values of 250 ms, 310 ms, 370 ms, 430 ms, 490 ms, and 550 ms, and the pitch contour realized with a rise of 0 Hz, 25 Hz, 50 Hz, or 75 Hz. These values were determined based on the results of Lo et al. (2019), that is, in a way that combining the two prosodic cues, duration and rising in pitch, would include the three contours found in the production experiment (see Table 2). Some of these manipulations are presented schematically in Figure 1. 5.2 Procedure Seventeen native speakers of Cantonese took part in the experiment. In each trial, after listening to the stimulus, the participant completed a three-alternative forced choice task, with each option exemplifying one of the three readings. For 230 7 Cantonese sentence-final particles & rhetorical questions Duration of SFP = 250 (ms) Duration of SFP = 550 (ms) 1.50 1.64 1.77 1.91 2.05 1.50 1.64 1.77 1.91 2.05 200 225 250 275 Time (s) Fundamental frequency (Hz) Figure 1: Schematic representation of some manipulations. instance, upon hearing an utterance like (31), with a prosodically manipulated sentence-final particle, the participant had to pick one of the three options shown in (32) that they thought matched the intonation of the utterance. (32) a. 講嘢嘅人 gong2je5ge3jan4 speaker 唔 m4 not 知道 zi1dou3 know 邊個 bin1go3 who 想 soeng2 want 飲 jam2 drink 咖啡 gaa3fe1 coffee (ISQ) ‘The speaker does not know who wants to drink coffee.’ b. 講嘢嘅人 gong2je5ge3jan4 speaker 認為 jing6wai4 think 冇人 mou5jan4 nobody 想 soeng2 want 飲 jam2 drink 咖啡 gaa3fe1 coffee (RQ−) ‘The speaker thinks nobody wants to drink coffee.’ c. 講嘢嘅人 gong2je5ge3jan4 speaker 已經 ji5ging1 already 知道 zi1dou3 know 邊個 bin1go3 who 想 soeng2 want 飲 jam2 drink 咖啡 gaa3fe1 coffee (RQ+) ‘The speaker already knows who wants to drink coffee.’ 5.3 Analysis Given that there were three possible responses for each trial, we employed a Bayesian multilevel multinomial logistic regression, with a random intercept for each participant but shared fixed effects for all participants (Koster & McElreath 2017). The reference level for the dependent variable was the informationseeking question response. Alongside the intercept, the predictor variables included duration and pitch rise, both of which were z-transformed before entering the analysis. The priors for all coefficients were a normal distribution with a mean of 0 and a standard deviation of 1. For the correlation structure of the 231 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung Dehé, Nicole & Bettina Braun. 2020. The intonation of information-seeking and rhetorical questions in Icelandic. Journal of Germanic Linguistics 32(1). 1–42. Farkas, Donka F. 2020. Canonical and non canonical questions. Ms. UC Santa Cruz / Princeton University, semanticsarchive.net. Farkas, Donka F. & Kim Bruce. 2010. On reacting to assertions and polar questions. Journal of Semantics 27. 81–118. DOI: 10.1093/jos/ffp010. Farkas, Donka F. & Floris Roelofsen. 2017. Division of labor in the interpretation of declaratives and interrogatives. Journal of Semantics 34(2). 237–289. DOI: 10.1093/jos/ffw012. Guerzoni, Elena. 2004. Even-NPIs in yes/no questions. Natural Language Semantics 12. 319–343. Gunlogson, Christine. 2003. True to form: Rising and falling declaratives as questions in English. New York: Routledge. Gussenhoven, Carlos. 2004. The phonology of tone and intonation. Cambridge: Cambridge University Press. Gussenhoven, Carlos & Aoju Chen. 2000. Universal and language-specific effects in the perception of question intonation. In 6th International Conference on Spoken Language Processing, ICSLP 2000. Beijing. Hamblin, Charles L. 1973. Questions in Montague English. Foundations of Language: International Journal of Language and Philosophy 10. 41–53. Han, Chung-Hye. 1998. Deriving the interpretation of rhetorical questions. In West Coast Conference of Formal Linguistics (WCCFL) 16, 237–253. Seattle. Han, Chung-Hye. 2002. Interpreting interrogatives as rhetorical questions. Lingua 112. 201–229. Jamieson, E. 2018. Questions, biases and ‘negation’: Evidence from Scots varieties. Edinburgh: University of Edinburgh. (Doctoral dissertation). Kharaman, Mariya, Manluolan Xu, Carsten Eulitz & Bettina Braun. 2019. The processing of prosodic cues to rhetorical question interpretation: Psycholinguistic and neurolinguistic evidence. In Interspeech 2019, 1218–1222. Kiss, Angelika & Roger Yu-Hsiang Lo. 2021. Two types of rhetorical questions: Support from Mandarin prosody. In Sinn und Bedeutung 26, 480–496. Koster, Jeremy & Richard McElreath. 2017. Multinomial analysis of behavior: Statistical methods. Behavioral Ecology and Sociobiology 71. 138. Lee, Wai-Sum. 2004. The effect of intonation on the citation tones in Cantonese. In First International Symposium on the Tonal Aspects of Languages, 107–110. Beijing. Lo, Roger Yu-Hsiang & Angelika Kiss. 2020. Durational and pitch marking of rhetorical wh-questions in Mandarin. In Proceedings of Speech Prosody 10, 429– 433. Tokyo. 238 7 Cantonese sentence-final particles & rhetorical questions Lo, Roger Yu-Hsiang, Angelika Kiss & Maxime Alexandra Tulling. 2019. The prosodic properties of the Cantonese sentence-final particles aa1 and aa3 in rhetorical wh-questions. In Sasha Calhoun, Paola Escudero, Marija Tabain & Paul Warren (eds.), 19th International Congress of Phonetic Sciences, 502–506. Melbourne. Luke, Kang Kwong. 1990. Utterance particles in Cantonese conversation. Amsterdam/Philadelphia: John Benjamins. Ma, Joan K-Y, Valter Ciocca & Tara L. Whitehill. 2006a. Effects of intonation on Cantonese lexical tones. Journal of the Acoustic Society of America 120(6). 3978– 3987. Ma, Joan K-Y, Valter Ciocca & Tara L. Whitehill. 2006b. Quantitative analysis of intonation patterns in statements and questions in Cantonese. In Proceedings of Speech Prosody 2006. Dresden. Matthews, Stephen & Virginia Yip. 2011. Cantonese: A comprehensive grammar. London: Routledge. Pierrehumbert, Janet & Julia Hirschberg. 1990. The meaning of intonational contours in the interpretation of discourse. In Philip R. Cohen, Jerry L. Morgan & Martha E. Pollack (eds.), Intentions in communication, 271–311. Cambridge: MIT Press. Prieto, Pilar & Paolo Roseano. 2021. The encoding of epistemic operations in two Romance languages: The interplay between intonation and discourse markers. Journal of Pragmatics 172. 146–163. DOI: 10.1016/j.pragma.2020.11.008. Roelofsen, Floris & Donka F. Farkas. 2015. Polarity particle responses as a window onto the interpretation of questions and assertions. Language 91(2). 359–414. Rohde, Hannah. 2006. Rhetorical questions as redundant interrogatives. San Diego Linguistics Papers 2. 134–168. https : / / escholarship . org / uc / item / 4xd7t5ww. Sadock, Jerrold M. 1971. Queclaratives. Papers from the Regional Meeting of the Chicago Linguistic Society (CLS) 7. 223–231. Sadock, Jerrold M. 1974. Towards a linguistic theory of speech acts. New York: Academic Press. Schaffer, Deborah. 2005. Can rhetorical questions function as retorts? Is the Pope Catholic? Journal of Pragmatics 37. 433–460. Schubiger, Maria. 1965. English intonation and German modal particles: A comparative study. Phonetica 12. 65–84. Stalnaker, Robert. 1978. Assertion. In Peter Cole (ed.), Syntax and semantics, vol. 9, 315–332. New York: Academic Press. Stenius, Erik. 1967. Mood and language-game. Synthèse 17. 254–274. DOI: 10.1007/ BF00485030. 239 Angelika Kiss, Roger Yu-Hsiang Lo & Justin R. Leung Sybesma, Rint & Boya Li. 2007. The dissection and structural mapping of Cantonese sentence final particles. Lingua 117. 1739–1783. Tsui, Rachel Ka-Ying & Shelley Xiuli Tong. 2018. An acoustic study of the effect of the interaction between intonation and lexical tones on the realization of Cantonese sentence-final particles. In Speech Prosody 9, 45–49. Wakefield, John C. 2012. A floating tone discourse morpheme: The English equivalent of Cantonese lo1. Lingua 122. 1739–1762. Wakefield, John C. 2019. Studies in Cantonese discourse. In Chris Shei (ed.), The Routledge handbook of Chinese discourse analysis, 569–581. London: Routledge. Xu, Bo Robert & Peggy Mok. 2012. Cross-linguistic perception of intonation by Mandarin and Cantonese listeners. In Speech Prosody 6, 99–102. Zahner, Katharina, Manluolan Xu, Yiya Chen, Nicole Dehé & Bettina Braun. 2020. The prosodic marking of rhetorical questions in Standard Chinese. In Speech Prosody 10, 389–393. 240