On the role of the Speaker’s beliefs in some biased questions

This short paper looks at two biased question types, Polar Rhetorical Questions (PRQs) and the so-called Negative Wh-Construction (NWHC). Although at first glance both could seem RQs because of the Speaker’s bias toward the negation of the proposition expressed by the sentence radical of the interrogative, it is argued here that the NWHC is not a subtype of RQs. Beliefs attributed to the Addressee play a crucial role in setting NWHCs and (polar) RQs apart.

1.1 The Negative Wh-Construction

The Negative Wh-Construction is a special question type: by its form it is a wh-interrogative, but by its function, it is a denial to some previous utterance.

Since when in B’s reaction bears emphatic stress and expresses that B does not believe the proposition that John is a vegetarian. Wh-interrogatives expressing such a move have been observed in a wide range of unrelated languages.

Although at a first glance, these questions look similar to Rhetorical Questions (RQs), the present paper argues that they are not rhetorical. According to Cheung (2008, 2009), the NWHC has a “negative rhetorical intrpretation”, by which he means that the at-issue meaning of a Negative Wh-Construction is a negative assertion. There are two discourse-related constraints that apply, both of which are conventional implicatures. One of them he called the ‘Conflicting View Condition’, which restricts the utterance of a NWHC to contexts where the Speaker is sure that her conversational partner believes the opposite of the proposition currently in question, and the other one being the so-called ‘Mis-Conclusion Condition’, which guarantees that NWHCs be used in contexts in which the Speaker thinks that her partner should have come to the same conclusion as the Speaker. The three meaning components are shown in (3), where p the proposition expressed by the previous utterance to which the NWHC is a denial.

Applying these to (1), upon uttering ‘Since when is John a vegetarian?’, the Speaker commits herself to ‘John is not a vegetarian’, and thinks that her Addressee (wrongly) believes that John is a vegetarian. Also, the Speaker thinks that the Addressee should have already known that John was not a vegetarian.

It has been noted that some NWHCs do not necessarily express propositional negation but they can target some metalinguistic issue raised by the latest move (Kiss, 2017). The present paper restricts itself to NWHCs that express propositional negation.

1.2 The markedness of NWHCs

NWHCs are thus biased questions, showing a marked form in several ways. As for their syntax, they resist any kind of embedding and the wh-phrase is restricted to a peripheral position. The examples in (4) could only be grammatical if the embedded sentence is pronounced as a quote, but not otherwise. Example (5) shows that a NWHC cannot serve as a sentential subject.

The peripheral position of the wh-phrase in a NWHC is seen even in wh-in-situ languages. The Cantonese example in (6) shows that unless the wh-phrase is above the ability modal wui, the utterance cannot be interpreted as a NWHC. (6a) and (6b) are genuine questions, (6c) is a NWHC.

Cheung has made the following observations: Languages differ in what wh-phrases can participate in a NWHC, if any, but typically only a small subset of wh-phrases is used, and they cannot be paraphrased. As for their semantics, the wh-phrases do not quantify over the domains canonically associated with them: where does not range over places, when does not range over times, etc. (Cheung analyzes them as wh-words that stand for possible worlds). Also, the wh-phrase in a NWHC can easily cooccur with a constituent that in a genuine question would serve as a congruent answer to it.

Another property contributing to markedness is that the wh-phrase in NWHCs can cooccur with predicates having temporal properties that would otherwise make them incompatible in genuine questions. Since when is normally ungrammatical with a telic predicate like decide to vote for Trump, and where is incompatible with stative properties like to be a bus.

Furthermore, the wh-expressions in NWHCs tend to be marked in form in some languages, which we could regard as another support for the claim that these wh-expressions are not the same semantically as the ones in genuine questions. Maybe they are not outright idiomatic, but they are somewhere between their ordinary meaning known from genuine questions and the special meaning seen in NWHCs. Examples of such wh-phrases are shown in (9):

English since when is close to the literal end of the continuum, but at the other end, we find examples such as ma dove ‘but where’ in Italian and kit^hõ, derived from kit^he in Punjabi.

Lastly, NWHCs have a falling final tune, which differentiates them from genuine questions, and the wh-phrase bears emphatic stress.

The above properties show that the relation of the sentence radical and the wh-phrase is different from how it is in genuine questions. The sentence radical of a NWHC is of the same type as the sentence radical of declaratives and polar questions, that is, they are full propositions. While the sentence radical of wh-interrogatives lacks information (Krifka, 2011), the one of NWHCs does not. Examples (7) and (8) prove that the wh-phrase does not come from the proposition as it is in genuine questions. In NWHCs, the wh-phrase seems to act as a discourse-related operator, which is supported by the fact that NWHCs cannot embed.

1.3 What this paper is about

Cheung assumes that the negative meaning of NWHCs arises from what he calls a “negative rhetorical interpretation”. He argues that NWHCs and RQs share two common traits: that in either case, no answer is expected, and that there is no such value in the quantification domain that could make the proposition true (Cheung, 2009, p. 317). Due to these properties, NWHCs are “sometimes felt to be like Rhetorical Questions”. But RQs can be answered and it is possible to find an answer in the domain canonically associated with the wh-phrase, therefore the two question types differ crucially. The example of an answered RQ in (10) illustrates these arguments.

According to Cheung, the rhetorical interpretation of a wh-question is determined pragmatically, because it depends on the availability of an obvious answer that all participants know. In the case of NWHCs, on the other hand, the Mis-Conclusion Condition exludes any possible answers; Cheung treats this condition as part of the semantics of NWHCs, hence the difference from RQs.

The present paper aims to spell out the difference between RQs and NWHCs by representing the beliefs of the Speaker. This is shown informally in section 2. In section 3, the system of Farkas and Roelofsen (2017) is presented, which offers a principled way to associate the semantic content and conventions of use of declaratives and polar interrogatives. In this framework, only those utterances can be accommodated that have a proposition-denoting sentence radical and have a so-called highlighted proposition. It will be shown that both PRQs and NWHCs contribute a highlighted proposition. Since this framework also cares about the Speaker’s attitude towards the truth of the proposition conveyed by some biased questions, the question whether PRQs and NWHCs should be accommodated in Farkas and Roelofsen’s framework becomes relevant.

2 Belief-constellations

2.1 What makes a question rhetorical?

Given Cheung’s claim that NWHCs have a ‘negative rhetorical interpretation’ but are not Rhetorical Questions themselves, it is worth looking at where this interpretation comes from. What makes a question rhetorical? There are two views found in the literature on how one could analyze RQs. One of them analyzes them as assertions (Han and Siegel, 1997; Han, 1998, 2002), and the other, as questions (Rohde, 2006; Caponigro and Sprouse, 2007). The assertion-like analysis is based on the fact that RQs pattern with assertions, rather than with questions, in some tests offered by Sadock (1974), some examples of which are shown in (11). A genuine question cannot follow after all, but an assertion can. On the other hand, by any chance can only appear in a genuine question, not in a RQ or in an assertion. What we see is that RQs pattern with assertions.

The semantic derivation at some point turns RQs into assertions of the opposite polarity, which is shown in (12). The job is done by an operator that turns the polarity of the sentence radical to the opposite (Han, 1998, 2002).

Since the polarity operator is set to a pragmatically available value, the expression is not a question any more, it becomes an assertion. One major advantage of the assertion-like analysis is that it also accounts for the distribution of Negative Polarity Items in RQs: it is the hidden operator that licenses them.

Caponigro and Sprouse (2007) and Rohde (2006) have pointed out some problems with this analysis, both in terms of empirical facts and theoretically. They noticed that RQs can be answered, while assertions cannot. Also, RQs can have answers that are positive, that is, non-negated propositions.

Rohde (2006) points to Schaffer’s (2005) RQs-as-retorts, another example of RQs not having an answer “of opposite polarity of what has been asked”, as Han claims is the case.

The RQ in (14aB) does not convey the negation of what has been asked, on the contrary. And neither does (14bB), strictly speaking: what it says is ‘I like school to the extent you like prison’, and the answer also signals that the question was a trivial one. RQs-as-retorts are special in several ways, the reader is referred to Schaffer (2005) for a description of them.

Furthermore, Asher and Reese (2007) have observed that Sadock’s tests can sometimes yield contradictory results. Consider the following example of a RQ, where after all (compatible with RQs and assertions only) and by any chance (compatible with genuine questions only) cooccur:

As it was shown above, after all should only introduce utterances that convey an assertion, while by any chance is associated with genuine questions, excluding assertions – Sadock’s tests are thus not fully reliable¹

As for the theoretical side of the assertion-like analysis, one cannot overlook the costliness of Han’s assumptions. First, the interpretation of interrogatives should branch into questions and assertions, from now on, despite the fact that they can have the same form (at least the same segmental material), as pointed out by Caponigro and Sprouse (2007). Also, Han assumes two sets of wh-words: interrogative and negative ones, yet we have hardly any reason based on morphology to do so.

The other view found in the literature, advocated by Caponigro and Sprouse (2007) and by Rohde (2006), pictures Rhetorical Questions as questions. On this account, it is the obviousness of the answer, and nothing else, that makes a question rhetorical. In other words, genuine information-seeking questions and RQs are the same semantically, although they differ pragmatically; what makes them different is the public beliefs the discourse participants are committed to. Once the speakers all believe the same answer to a given question, it can be asked felicitously as a RQ.

What makes a question rhetorical in Caponigro and Sprouse’s analysis is formulated in (16):

Treating RQs as questions has several advantages compared to the assertion-like analysis: There is no need to posit two sets of wh-words, the facts about answerability are explained and Sadock’s tests are not worrying either, knowing that they are not entirely reliable. The major shortcoming of the question-like analysis is that it does not offer an answer to why NPIs can occur in RQs. This problem is not treated here, but the reader is referred to the work of Guerzoni (2004) and Abels (2003) who offer some answer to it. Despite this shortcoming, in this paper, RQs are analyzed as questions to which the answers are perceived by the Speaker as obvious.

2.2 The role of discourse participants’ beliefs

A RQ suggesting a negative answer is similar to a NWHC in an important respect: the Speaker believes that the answer to her question is the empty set. What RQs and NWHCs differ in is the belief the Speaker attributes to the Addressee about the truth of the proposition in question. Following Caponigro and Sprouse, in the case of a RQ, the Speaker attributes a positive or negative belief to herself about the proposition in question p if the RQ in question is a polar one. If the RQ is a wh-interrogative, this negative belief means that according to the Speaker, there is no value that makes the sentence radical a true proposition. At the same time, the Speaker attributes the same belief on the proposition p to her Addressee, which is either the sentence radical of a PRQ or a proposition that serves as a true answer to the wh-RQ. This condition holds whether it is a RQ awaiting a positive or a negative answer.

In the case of a NWHC, recall that there are some discourse-related constraints, namely the Conflicting View Condition and the Mis-Conclusion Condition. These guarantee that a NWHC is pronounced felicitously only if the Speaker and the Addressee have opposing beliefs about p. Upon reacting to the previous utterance expressing the proposition p, a NWHC expresses ¬p.

Let S^S represent the Speaker’s own belief about the proposition in question, and let S^A be the belief attributed by the Speaker to the Hearer. The polarity of the belief about the truth of p is marked by +, – and neut for positive, negative and neutral stances on it. In this case, RQs with negative answers, RQs with positive answers and NWHCs pertain to three different constellations of beliefs. Let us represent a certain belief-constellation as an ordered pair of Speaker’s beliefs, the first element being S’s belief about p and the second, what S attributes to A as a belief about p.

Both RQs and NWHCs are biased questions, since in both cases, the Speaker has a bias toward the proposition she is questioning, and in addition, she also attributes a bias to the Addressee. The similarity between RQs and NWHCs is due to the Speaker’s negative bias S^S- in both cases. But the Addressee is attributed different things in both cases: RQs with a negative answer require the Addressee’s belief to be negative as well, while NWHCs are felicitous only if the Addressee’s belief on the subject matter p is of opposite polarity, that is, positive.

These properties are summarized in the following table, where S stands for Speaker, and A, for the belief attributed to the Addressee by S:


S’s belief about A →	S^A-	S^A-neut	S^A+
S’s own belief

S^S-	RQ^-	accommodate as a RQ	NWHC

S^S-neut	n/a	genuine question	rising declarative

S^S+	n/a	accommodate as a RQ	RQ⁺

Table 1: Speaker’s beliefs about the proposition in question and about Addressee’s beliefs on it.

The two dimensions of the table show the possible beliefs of the Speaker about the truth of the proposition p² . The header column shows that the Speaker can believe that p is not true (S^S-), that p is true (S^S+), or she can have a neutral stance on it (S^S-neut). The header row shows three possible beliefs the Speaker can attribute to the Addressee: she can think A agrees with p (S^A+), that A disagrees with p (S^A-), or it is possible that the Speaker has no reason to attribute any bias to A (S^A-neut). Attributing a neutral stance to A does not mean to believe that A does not know whether p is true, it only means that the Speaker is unable to attribute a positive or a negative stance to A about p, which does not exclude that A does have one. S^S-neut, on the other hand, should be read off as “S cannot attribute neither a positive, neither a negative attitude to S”, that is, to herself, which means that the Speaker does not know whether p is true or not.

If we adopt Caponigro and Sprouse’s view on the “felicity conditions” of RQs, then both members of the ordered pair should show the same (non-neutral) value. To illustrate it, consider the following example of a RQ awaiting a negative answer.

Here, the proposition in question p would be the sentence radical of the RQ, chickens have lips. The answer to this question is negative and this fact is obvious enough so that we can rely on that the Addressee also knows this. The belief constellation of the Speaker of (20) is therefore ⟨S^S-,S^A-⟩.

The proposition in question here is composed of applying the commonly known answer to the question radical (in the terminology of Krifka (2017)): λx[x helped Luca when he was in trouble](Mina). The proposition in question in this case would be ‘Mina helped Luca when he was in trouble’. The Speaker believes this proposition is true and expects her Addressee to also believe this to be the case, the belief constellation is thus ⟨S^S+,S^A+⟩.

As for the rest of the cases in Table 1, a Speaker can use a RQ felicitously even if she does not believe the Addressee to be biased towards the same answer her question suggests because as Abels (2003) noted, a cooperative discourse participant can accommodate presuppositions arising from RQs. This represents the notion of “one-sided mutual belief”, the importance of which is highlighted by Pierrehumbert and Hirschberg (1990). Belief constellations that have a neutral value for S’s own belief about p are genuine questions in case the belief attributed to A is not biased either: ⟨S^S-neut,S^A-neut⟩. If A is biased towards the positive answer, however, it is a case of a rising declarative (Gunlogson, 2001). ⟨S^S-neut,S^A+⟩ represents cases of questions that in English are syntactically declarative but have a rising intonation. Rising declaratives fail to commit the Speaker to the truth of the sentence radical, yet they are biased, namely their Speaker attributes a certain proposition to the Addressee (Gunlogson, 2001, p. 4). If this is indeed the constellation of beliefs a rising declarative expresses, then it can be adequately captured by the ordered pair of Speaker’s beliefs ⟨S^S-neut,S^A+⟩.

In sum, Rhetorical Questions and the Negative Wh-Construction are similar in an important respect, namely that the Speaker’s negative stance on a proposition in question p (S^S-) is part of the felicity conditions of both. What sets them apart is the belief the Speaker attributes to the Addressee.

2.3 Intonation

It has been noted in the literature that Rhetorical Questions typically have a low boundary tone L% at least in English (see Banuazizi and Creswell, 1999 and the references therein). Banuazizi and Creswell (1999) found that RQs are far less likely to rise than genuine polar questions. They explain this tendency in terms of the Speaker’s beliefs: the Speaker has the answer to her own question in her mind, which is “either perfectly obvious or perfectly obviously unknowable”, thus the Speaker definitely does not need an answer. Final falls are associated with assertiveness, that is, they do not need to be interpreted “with respect to subsequent utterances” (Pierrehumbert and Hirschberg, 1990).

Although no recordings or measurements have been made of NWHCs yet, it is a rather salient property of them that they have a falling, rather than rising contour at their end. Pierrehumbert and Hirschberg (1990) state that “final lowering reflects the degree of “finality” of an utterance, the more final lowering, the more the sense that an utterance “completes” a topic” (p. 279). Although Banuazizi and Creswell acknowledge that one cannot draw a direct correlation between the meaning/function of questions and their intonation, they came to the conclusion that discourse functions do correlate with final contour – at least to some degree of predictability. What is of particular interest here is that they found a phenomenon that vaguely parallels the epistemic stance of the Speaker with respect to the proposition in question.

3 Semantic content and conventions of use

Farkas and Roelofsen (2017) suggest a framework that allows that semantic content and conventional discourse effects divide the labor when interpreting declaratives and polar interrogatives. Besides information-seeking questions, they also look at some special questions such as rising declaratives, claiming that their framework can host all question types, biased and neutral. It is argued here, that we have reasons to accommodate RQs and NWHCs in their system, and if so, some extensions are needed.

The authors claim that the difference in the interpretation of assertions and polar interrogatives follows from their semantic differences while the same convention of use is assigned to both. Marked utterance types, for example rising declaratives are also subject to some special discourse effects, which is due to how strongly the Speaker believes that the highlighted proposition (the sentence radical) is true. This they call the credence level, and it becomes important whenever special effects arise, and special effects arise due to markedness.

The authors characterize the basic discourse context in terms of the set of discourse participants, the table, which is a stack of propositions that have been raised as issues but have not yet been accepted into the common ground (Farkas and Bruce, 2010), and the set of commitments, which maps discourse participants to the set of propositions that cover all their public beliefs (to their commitment set, see Stalnaker (1978)).

Farkas and Roelofsen characterize the semantics of declaratives and interrogatives in terms of clause type markers. The presence of the clause type markers dec or int are signaled in English by word order, and the clause type markers closed and open are signaled by falling and rising intonation, respectively. The two kinds of markers together determine whether the sentence is inquisitive or not (Ciardelli et al., 2013). The sentence radical is the argument of the dec/int marker, which determines whether it is inquisitive or not, and the resulting expression then combines with the closed/open marker, which can have a further effect on its inquisitiveness.

In the case of an assertion, the clause type marker dec makes any sentence radical non-inquisitive (marked by !), and if it is also closed, the utterance ends up non-inquisitive, that is, it is informative and does not need further information to be completed. On the other hand, polar interrogatives expressing an information-seeking question are typed int and open, marked by ?. Marked sentence types typically have both ! and ?, as it is exemplified by the case of rising declaratives.

Since in this case a declarative combines with open, both the highlighted alternative P and its complement are added to the table, and the sentence becomes inquisitive. A proposition is equivalent to a set of worlds: adding the proposition together with its complement results in adding the whole set of worlds W to the table, which is not an informative move.

In case of an assertion, ϕ equals to a proposition, but in case of a polar interrogative, ϕ stands for both the highlighted alternative and its complement, as shown in (27). Thus, in the latter case, the entire set of worlds W is added to commitments(x), as it becomes an inquisitive proposition, resulting in a trivial commitment, requiring an answer.

In case of a genuine question, the answer is supposed to reduce the set of worlds in commitments(x). But in marked sentences, some special discourse effects may arise. Between any two forms that have the same semantic content, the one that is formally more complex, and therefore “less likely to ensure communicative success” is considered marked (p. 263). Assuming that both segmental and suprasegmental material counts as ‘form’, biased questions are all marked, as expected. Tag interrogatives have a longer form than polar interrogatives without a tag, and rising declaratives, RQs and NWHCs are marked because of their intonation. And NWHCs, in addition, also have peculiar syntactic properties that in their genuine counterparts would cause ungrammaticality.

Marked sentences are subject to the same basic convention as unmarked ones, but in addition, they carry some extra information that is signaled by their non-minimal form. The extra information, according to Farkas and Roelofsen, concerns the level of credence the Speaker has in the truth of the highlighted proposition. In characterizing the discourse context, the authors make use of a set they called evidence(x), which stands for the set of propositions, such that speaker x has access to evidence to believe them (evidenced possibilities as Farkas and Roelofsen call it). In addition to signalling that the Speaker has evidence to the truth of the highlighted proposition, she also signals the level of confidence she has in the truth of it. She can have high credence, as in the case of uttering an assertion, or low credence, when uttering a rising declarative, or in yet other cases, moderate or zero credence. The set evidence is thus made up of ordered pairs ⟨p, i⟩ such that p is a proposition the Speaker has evidence to believe and i is an interval of credence levels. In the case of a rising declarative, the special effect equals to a low level of credence on the part of the Speaker.

Farkas and Roelofsen’s system thus determines whether a sentence is inquisitive or not by using clause type markers. These sentences are then subject to the same basic convention, namely that the inquisitive or non-inquisitive expression is added to the table and the corresponding set of worlds is added to the set of commitments. Thirdly, if the sentence is marked, some special effects are postulated, which share one characteristic: they all say something about the level of the Speaker’s confidence in the truth of the highlighted proposition. However, the authors admit that special effects are assigned to each sentence form individually, but not according to a principle.

3.1 PRQs and the NWHC in Farkas and Roelofsen’s system

Farkas and Roelofsen do not consider the case of RQs or NWHCs. They mention that RQs are used to underline some proposition already present in the common ground, but this property of them belongs to pragmatics. However, both RQs and NWHCs are marked utterances based on the author’s measures, because of the mismatch between their form (an interrogative) and what they convey (an assertion), which makes them “more prone to misinterpretation”. Intonation also alters the form so that it becomes more marked, as RQs are more likely to have a sentence-final falling intonation (Banuazizi and Creswell, 1999), and NWHCs have a similar tendency.

There are some other good reasons to treat PRQs and NWHCs in the same system, as they both have sentence radicals that are full propositions. This is obvious for PRQs, but less so for NWHCs. In section 1.2, it was shown that the wh-expression cannot stand for a constituent that comes from the clause, thus NWHCs are made up of a wh-phrase independent from the clause (Cheung, 2009), and a proposition-denoting sentence radical.

Also, these sentence radicals pass the tests of highlighted propositions suggested by Farkas and Roelofsen, namely they can serve as antecedents for anaphoric expressions. Such tests include expressions like yes, no, if so and otherwise. Just as the sentence radicals of genuine polar questions, rising declaratives, and tag questions can be referred to by such expressions, as shown in (29), so can the sentence radicals of NWHCs (30b) and PRQs (30c), too, setting them apart from wh-interrogatives, whether they are genuine or rhetorical (30a).

The same contrast is shown in English with (31): let alone genuine wh-interrogatives, all other question types under discussion have highlighted propositions that it can refer back to.

Given that the PRQs and NWHCs seem to contribute highlighted propositions to the discourse just like assertions, polar interrogatives, rising declaratives and tag questions, there is reason to accommodate them in the same framework.

If int contributes inquisitiveness and closed encodes falling intonation, a PRQ or a NWHC should then have the following clause type markers:

In the case of a PRQ, the argument of int is a proposition which needs int to become inquisitive. Since closed is semantically vacuous (see (24c), the result is an inquisitive expression, as expected. The interpretation is shown in (33b).

The conventional discourse effects of a PRQ are the same as with a genuine polar question. But because of its markedness, it is expected to have special effects. If the highlighted alternative is the sentence radical, the special effect should be a strong disbelief in it. This, according to Farkas and Roelofsen, is captured by a zero credence level, which can mean that the Speaker is unbiased or that she is against the truth of the highlighted proposition (p. 256).

The problem that arises here is that the basic and special conventional discourse effects of a PRQ is not any different from a genuine polar question. Since in the case of a PRQ awaiting a negative answer, the Speaker believes the opposite of the highlighted proposition, saying that she has a zero credence level is not enough. It is suggested that the label contrary be used in this case, and let zero be saved for unbiased questions.

Now, the case of NWHCs is peculiar and it is well possible that they cannot be accommodated in Farkas and Roelofsen’s system at all. In what follows, we give a first look at the possibilities we have.

Recall that NWHCs differ from ordinary wh-interrogatives because their denotation is a set made up of full propositions, which is not necessarily the case with genuine wh-interrogatives; and because they contribute a highlighted proposition, unlike genuine wh-interrogatives. Also, the wh-phrase in a NWHC does not range over entities it is canonically associated with, e.g. since when in (1B) does not range over times (Cheung, 2008, 2009), which is corroborated by the fact that the segmental material of wh-phrases in NWHCs is slightly altered in some languages.

If (36a) is pronounced as a NWHC, we cannot use the variable t standing for timepoints to create the partition on W, because the wh-phrase is no longer associated with its canonical domain, as it was shown in (7). And even if we decide that a Hamblin-set is put on the table, it is not clear which proposition to choose from the table as the highlighted one.

Otherwise, we can say that the highlighted proposition is the sentence radical of the NWHC and put it on the table along with its complement generated by int, as a result of which we get the following basic conventional discourse effects:

But (37) is not a fortunate representation either, because we represent a wh-interrogative (though not a canonical one) the same way we represent a polar interrogative. Until a better way of representing the conventional discourse effects of NWHCs is found, we let (37) be the provisional representation.

3.2 Special effects

As for the special effects of NWHCs, the Speaker is expected to believe that the highlighted proposition is false, that is, she will be assigned the credence level contrary, as in the case of PRQs awaiting a negative answer.

Finally, if PRQs and NWHCs have the same basic and special conventional discourse effects, cf. (34), (35), (37) and (38), then they need to be further distinguished. It was suggested in section 2 that RQs awaiting a negative answer and NWHCs both share the property that their Speaker disagrees with some proposition p, possibly conveyed by the latest move, which is also the sentence radical of the sentence. It was also shown that NWHCs are not RQs, because the Speaker attributes different beliefs to her Addressee in both cases. Namely, in the case of RQs, the Speaker believes and/or conveys that her Addressee has (or should have) the same belief on the truth of p, while in the case of NWHCs, the Speaker and Addressee have opposing beliefs. If so, the special effects could host an extra slot for the belief attributed to the Addressee.

Although representing both the Speaker’s belief and the belief attributed by the Speaker to the Addressee is more costly, as it complicates the representation, there seems to be no other way to capture the difference between NWHCs and RQs, at least given the premises this paper builds on.

There are various views in the literature on how it comes about that RQs convey their speakers’ bias (which is an assertion): it has been attributed to semantics (Han, 2002) and to pragmatics (Caponigro and Sprouse, 2007). As for NWHCs, Cheung argues that the special effect described here, which he called the Conflicting View Condition, is a conventional implicature⁴ . His claim is not inconsistent with the arguments of this paper. What we stated here is that PRQs and NWHCs, being marked utterances, have a highlighted proposition, based on the division of labor principle in (22), therefore they must have special effects. These special effects are present by virtue of the marked form, which means they are encoded, and being encoded is one of the tests Potts (2015) uses to tell conventional implicatures from other kinds of meaning. There are other tests though, and it is outside the scope of this paper to determine where these special effects belong, although it is not a minor issue.

Farkas and Roelofsen suggest that intonation and credence levels show a connection in marked sentence forms. The generalization they make is the following: the lower the credence level in the highlighted alternative, the less likely the utterance ends with a falling tune; and the higher the credence level, the less likely the utterance involves a final rise (p. 272). Now, if NWHCs have the special discourse effect shown in (38), in favor of which some supporting facts have been shown, then this generalization does not hold. In (1B), because of the falling final tune, the Speaker should strongly believe the truth of the sentence radical ‘John is a vegetarian’, which is just the opposite of how it actually is. Thus, as a next step, this prediction should be modified to possibly accommodate the belief attributed by the Speaker to the Addressee as well.

Lastly, a remark should be made about the status of RQs. What was considered here as special effect, Farkas and Roelofsen considers pragmatics. Undoubtedly, not all RQs are marked the way suggested here. Banuazizi and Creswell (1999) observed a tendency for RQs to have a sentence-final falling intonation, but by no means is this always the case for (English) PRQs. What is more, as Farkas and Roelofsen claims, genuine polar interrogatives can have a sentence-final fall, too. If so, it raises the question of how marked RQs are at all, and in what ways.

4 Conclusion

This short paper has presented some arguments for the importance of belief-constellations, that is, for representing a more elaborate picture of the Speaker’s own belief and percieved beliefs. The advantage of doing so is seen in a commitment-based account such as the one of Farkas and Roelofsen (2017), that aims to host sentences having a highlighted alternative. PRQs with a negative answer and NWHCs both have such alternatives, so there is reason to consider their place in such a model. Fitting them into Farkas and Roelofsen’s account raises questions; to some of them, the paper suggested an answer.

Acknowledgements

I am thankful to Michela Ippolito and Guillaume Thomas for the insights they gave while I was working on the Negative Wh-Construction earlier. I am also thankful for Beáta Gyuris and Lisa Cheng for having sparked my interest in the semantics of questions. But most importantly, I am thankful for László Kálmán, for having sparked my interest in semantics itself.

References

Abels, Klaus. 2003. Who Gives a Damn about Minimizers in Questions? In Proceedings of SALT XIII, ed. R. Young and Y. Zhou, 1–18. Ithaca, NY: Cornell University.

Asher, Nicholas, and Brian Reese. 2007. Intonation and discourse: Biased questions. In Working Papers of the SFB632, Interdisciplinary Studies on Information Structure (ISIS), ed. S. Ishihara, S. Jannedy, and A. Schwarz, volume 8, 1–38. Potsdam: Universitätsverlag Potsdam.

Banuazizi, Atissa, and Cassandre Creswell. 1999. Is that a real question? Final rises, final falls and discourse function in yes-no question intonation. In CLS 35, 1–14.

Caponigro, Ivano, and Jon Sprouse. 2007. Rhetorical questions as questions. In Proceedings of Sinn und Bedeutung, ed. E. Puig-Waldmüller, 121–133. Barcelona: Universitat Pompeu Fabra.

Cheung, Lawrence Yam-Leung. 2008. The Negative Wh-Construction. Doctoral Dissertation, UCLA.

Cheung, Lawrence Yam-Leung. 2009. Negative wh-construction and its semantic properties. Journal of East Asian Linguistics 18:297–321.

Ciardelli, I., J. Groenendijk, and F. Roelofsen. 2013. Inquisitive semantics: a new notion of meaning. Language and Linguistic Compass 7:459–476.

Farkas, Donka F., and Kim B. Bruce. 2010. On Reacting to Assertions and Polar Questions. Journal of Semantics 27:81–118.

Farkas, Donka F., and Floris Roelofsen. 2017. Division of Labor in the Interpretation of Declaratives and Interrogatives. Journal of Semantics 34:237–289.

Guerzoni, Elena. 2004. Even-NPIs in Yes/No questions. Natural Language Semantics 12:319–343.

Gunlogson, Christine. 2001. True to form: Rising and falling declaratives as questions in English. Doctoral Dissertation, University of California Santa Cruz.

Gyuris, Beáta. 2017. New perspectives on bias in polar questions: a study of Hungarian -e. International Review of Pragmatics 9:1–50.

Han, Chung-Hye. 1998. Deriving the Interpretation of Rhetorical Questions. In Proceedings of WCCFL 16, 237–253. Stanford: CSLI Publications.

Han, Chung-Hye. 2002. Interpreting interrogatives as rhetorical questions. Lingua 112:201–229.

Han, Chung-Hye, and Laura Siegel. 1997. Syntactic and Semantic Conditions on NPI Licensing in Questions. In Proceedings of WCCFL 15, 177–191. Stanford: CSLI Publications.

Kiss, Angelika. 2017. Meta-conversational since when-questions and the common ground. In Proceedings of the 21st Workshop on the Semantics and Pragmatics of Dialogue, ed. Volha Petukhova and Ye Tian, 38–47. Saarbrücken.

Krifka, Manfred. 2011. Questions. In Semantics. An International Handbook, ed. Claudia Maienborn, Klaus von Heusinger, and Paul Portner, 1742–1785. De Gruyter Mouton.

Krifka, Manfred. 2017. Negated polarity questions as denegations of assertions. In Contrastiveness in Information Structure, Alternatives and Scalar Implicatures, ed. Chungmin Lee, Ferenc Kiefer, and Manfred Krifka, 359–398. Springer.

Pierrehumbert, Janet, and Julia Hirschberg. 1990. The Meaning of Intonational Contours in Discourse. In Intentions in Communication, ed. Philip R. Cohen, Jerry Morgan, and Martha E. Pollack, 271–312. Cambridge, MA: MIT Press.

Potts, Christopher. 2015. Presupposition and implicature. In The Handbook of Contemporary Semantic Theory, ed. Shalom Lappin and Chris Fox, 168–202. Wiley Blackwell.

Rohde, Hannah. 2006. Rhetorical questions as redundant interrogatives. San Diego Linguistics Papers 2:134–168.

Sadock, J. 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.

Stalnaker, Robert C. 1978. Assertion. In Pragmatics, ed. Peter Cole, volume 9 of Syntax and Semantics, 315–332. New York: Academic Press.