Yes-no questions are those questions whose answer is either “yes” or “no” (see (1) for a Sardinian example). The fact that they are restricted to just two possible answers explains why they are also called “polar questions”. They contrast with wh-questions, which are questions asking about the word or phrase the wh-word replaces, as in (2) for Sardinian. The mechanisms used to mark polar questions in natural languages are diverse. According to Dryer (2013), they can be marked by means of (a) a question particle added to a declarative sentence to indicate that it is a question, as in (3) for Maybrat (Dol, 1999); (b) distinct interrogative verbal morphology (see (4) for Tunica; Haas, 1940); (c) a combination of both question particle and interrogative verbal morphology, as in (5) for Pirahã (Everett, 1986); (d) different word order (see (6) for German); (e) the absence of morphemes used in declaratives (see (7) for Zayse; Hayward, 1990); or (f) a distinct intonation pattern. This cross-linguistic marking of polar questions is mainly restricted to neutral questions, i.e., questions for which the speaker has no particular expectations about the answer. However, we know that the intention behind the act of asking can be more than just a mechanism to solicit information, but may instead reflect a need to confirm assumptions, formulate an offer, express surprise, etc.

In recent years a considerable amount of research has been devoted to the interface between semantics-pragmatics and prosody in polar questions for different languages dealing with information structure and/or degree of certainty about the truth value of the proposition (Grice & Savino 1997, 2003, for Italian; Haan, 2001, for Dutch; Kügler, 2003, for German; Santos & Mata, 2008, for Portuguese; Vanrell, Mascaró, Torres-Tamarit, & Prieto, 2013, for Catalan); speaker bias (Asher & Reese, 2007; Hwang & Ito, 2014, for Japanese; Armstrong, in press, for Spanish); information source (Hara & Kawahara, 2012, for Japanese; Vanrell, Armstrong, & Prieto, 2014, for Catalan) and/or requests (Álvarez & Blondet, 2003; Estrella-Santos, 2007; Orozco Vaca, 2008, 2010, for Spanish; Astruc, Vanrell, & Prieto, in press; Nadeu & Prieto, 2011, for Catalan); and incredulity or disbelief (Lee, 2010, for Spanish; Crespo-Sendra, 2011, for Catalan). These studies tend to concentrate on the intonational patterns used to convey specific meanings but they often disregard the lexico-syntactic structure itself or how these tonal events interact with lexicon and syntax. Additionally, as pointed out by Armstrong (in press) most of these works use coarse-grained distinctions like information-seeking questions or unbiased questions vs. confirmation-seeking questions or biased questions, which do not always “allow us to capture different points on the epistemic gradient that might be candidates for intonational encoding”. In this paper we will make an attempt to encompass the aforementioned aspects and reconcile both the lexico-syntactic and the prosodic perspective.

Romance languages are often described as marking questions by means of a distinct intonational pattern. This may be attributable to the fact that most Romance languages (with the exception of French) are pro-drop, lack an auxiliary verb (compared to some other European languages) and present a more free word order. In fact, a variety of question-marking strategies can be seen in different Romance languages including word order (Rigau, 2002, for Catalan; Brandi & Cordin, 1989; Poletto, 1991; Renzi & Vanelli, 1982; Rizzi, 1986, for Friulian and Rhaeto-Romance and Gallo-Italic languages; Vanrell & Fernández-Soriano, 2013, 2014, for different Catalan and Spanish varieties); question particles (Prieto & Rigau, 2007, for Catalan; Fossat, 2006; Karenova, 2006; Pusch, 2000, for Gascon Occitan); and interrogative verbal morphology (Lusini, 2013, p. 38; Manzini & Savoia, 2005, p. 377, for the dialect of Cortemilia, Northern Italy). Sardinian is a Romance language that marks questions using three of these strategies, namely, question particles, word order and intonation.1 Sardinian polar questions can be headed by particles such as a (e.g., A benis a jocare chin mecus? ‘Are you coming to play with me?’; Jones, 1993), can also present constituent fronting (Mandicattu as? ‘Have you eaten?’, Ello, inoche ses? ‘Are you here?’, Mortu in s’ispidale est? ‘Did he die in the hospital?’; see Remberger, 2010) and can be characterized by different intonational patterns (¡H+L* L%, ¡H*+L L% in the Sard_ToBI system; see Contini, 1984; Schirru, 1982, 1992a, 1992b; and Vanrell et al., in press). Regarding constituent fronting, it has been characterized as a common marker in polar questions without being part of a question-formation process (Jones, 1993) and as a strategy for marking positive focus often with an emphatic value (i.e., the speaker expects it to be true that p; Remberger, 2010). On the other hand, the particle a has been described as being predominantly (but not exclusively) used in questions which are to be interpreted as requests, invitations and offers (Jones, 1993). For Jones (1993), the meaning of particle a is that of a politeness marker indicating the speaker’s willingness to accept denial or acceptance. Yet for Remberger (2010), there is no clear semantic difference between focus fronting and the marker a: “yes/no questions with focus fronting as well as those marked by a (which is clearly encoding positive polarity) will be treated as marked for positive polarity” (2010, p. 571). Floricic (2009) discusses the possibility that the particle a is a clitic, since it can appear as part of a clitic cluster with the partitive nde (e.g., Frutta, a nde cheres? ‘Fruit, do you want any?’; 2009, p. 132) and presents a series of distributional restrictions in that the particle a precedes the proposition that is questioned and this causes extra-sentential elements like vocatives or thematic constituents to be left out of the scope of the particle a (see (8)–(10)). However, ultimately Floricic (2009) discards this hypothesis by providing counterexamples like that seen in (11), in which the particle a bears the main stress of the sentence, and finally concludes that the function of the particle a is that of assigning “focal value to the content in its scope” (2009, p. 146). As for intonation, the ¡H+L* L% pattern tends to be related to information-seeking yes-no questions or questions with unspecified evidential/epistemic bias (Vanrell, Ballone, Schirru, & Prieto, 2014; Vanrell et al., in press), whereas the pattern ¡H*+L L% is related to confirmation questions, including echo yes-no questions (Vanrell et al., in press) or biased yes-no questions in general (Vanrell, Ballone, et al., 2014). Interestingly, the falling pattern generally appears with yes-no questions introduced by the particle a (e.g., A times? ‘Are you scared?’) or yes-no questions with no specific lexical/syntactic marker (e.g., A s’abba? ‘To the water?’). By contrast, yes-no questions presenting constituent fronting (e.g., Bidu l’as? ‘Have you seen him?’) tend to be characterized by a rising-falling intonation pattern (Contini, 1984).

In this paper we seek to refine the analysis presented in Vanrell et al. (in press) for the ¡H+L* L% and ¡H*+L L% tonal events associated with unbiased and biased polar questions respectively. We also aim to assess whether there exists a one-to-one correspondence between lexico-syntactic and intonational structure as is defended in Contini (1984) and Vanrell, Ballone, et al. (2014). According to these studies, the prosodic structure of a polar question is determined by its lexico-syntactic structure. Finally, we hope to shed some light on the specific meanings related to the linguistic markers found in Sardinian polar questions. Specifically, since question particles have been shown to mark not only questions but also different types of speaker bias in languages like Japanese (Hara, 2013; Sudo, 2013), we want to determine whether there is some sort of relationship between the different linguistic markers found in Sardinian polar questions and speaker bias.

The final corpus consists of 1125 sentences (Ìttiri n = 30; Ottieri n = 42; Puttumajore n = 40).

We will first examine the general results and then have a look at the data after separating it into the three different bias conditions (neutral in Section a, epistemic condition in Section b and evidential condition in Section c).

Figure 2 shows the incidence of each lexico-syntactic strategy (particle a, constituent fronting, negation, neutral word order6 and other) and each intonation pattern (¡H+L* L%, ¡H*+L L%, L+H* L% and other) in the general data. As can be seen, the preferred strategy used by the Logudorese speakers is the combination of constituent fronting with the ¡H*+L L% pattern (e.g., Ebbe’, cussu lìberu , battidu mi l’as? ‘So, this book, have you brought it to me?’). It appears in 45 instances, which represents 34% of the data. At some distance from this pattern, we find the same intonational pattern but with neutral word order (e.g., Ti nde ses ammentada, a mi comporare sas caramellas in farmacia? ‘Have you remembered it, to buy the cough drops at the pharmacy?’), which is used in 14 out of 112 cases. Less frequent strategies are the use of the ¡H*+L L% pattern with negation (e.g., Tando, frùtture e birdura, no nd bendides pius? ‘So, fruit and vegetables, you don’t sell them anymore?’) or with the particle a (e.g., A la tenes, una caramella? ‘Do you have it, a cough drop?’).

The examples labeled as displaying “other” lexico-syntactic strategies correspond to syntactic structures less common in the literature such as disjunctive polar questions (e.g., A bi nd’at o non bi nd’at, de bagna? ‘Is there tomato sauce or is there no tomato sauce?’) or questions with external adverbials of the type how come (cando mai in Sardinian; e.g., Ma cando mai ti as mandigadu tottas sas caramellas? ‘How come you have eaten all the candies?’).

This study set out to refine the analysis presented in Vanrell, Ballone, et al. (2014, in press) for the ¡H+L* L% and ¡H*+L L% nuclear configurations associated with unbiased and biased polar questions respectively. We also wanted to assess whether the correspondence between lexico-syntactic and intonational structure proposed in Contini (1984) and Vanrell et al. (in press) was supported by our data. Finally, we sought to provide some sort of explanation for the distribution of the different linguistic markers found in Sardinian polar questions. To this end, a production experiment was designed aiming to elicit polar questions with different bias and polarity conditions by means of the DCT methodology. The collected data were prosodically and syntactically annotated and then analyzed.

The results show a clear preference for the particle a in the neutral condition (e.g., A nde tenes, tumattas? ‘Do you have any, tomatoes?’), but two different intonational patterns can be found with this lexico-syntactic strategy: the ¡H+L* L% pattern (see Figure 4) and the ¡H*+L L% pattern (see Figure 5). If we take a closer look at the data, we realize that most of the ¡H*+L L% tokens were produced in the pragmatic situation in which the speaker meets a friend and while they are talking the speaker realizes that it might be possible for her to offer tomatoes to her friend, since she has plenty of them at home. The question was supposed to determine whether the interlocutor already had tomatoes (and therefore would not need any more), but the speaker could also be simply offering tomatoes. According to Jones (1993) and as also noted in the Introduction, the particle a can be related to questions which are to be interpreted as requests, invitations and offers, as well as to true polar questions. In addition, offers in Algherese (a Catalan variety spoken in northern Sardinia which exists in close contact with Logudorese Sardinian) can be headed by the particle a and tend to present the ¡H*+L L% pattern described above (e.g., A véns amb mi que te convid calqui cosa? ‘Do you want to come with me and I’ll buy you something?’; Vanrell, Roseano, & Cabré, 2013, p. 167). Interestingly, the particle a comes from the Latin aut ‘or’ which has an exclusive character compared to vel ‘or’, which had an inclusive meaning (Remberger, 2010, p. 570). According to Remberger (2010), this particle in origin was a former exclusive alternative operator similar to the Polish interrogative particle czy. A tentative hypothesis could be that the particle a is a politeness marker (already proposed by Jones, 1993) that mitigates the request for information in the case of true polar questions and conveys the speaker’s willingness to accept denial or acceptance in offers. That would explain why this particle fits well in both contexts, a neutral context and that of an offer/invitation.7 As for intonation, the ¡H+L* L% pattern (see Figure 4) could be marking “lack of bias” (we will explain later why it can appear in contexts with negative bias). On the other hand, the ¡H*+L L% pattern (see Figure 5) would be used to express bias based on beliefs, expectations, word knowledge or information that has become available in the discourse context. In the case of offers expressed through the combination of the particle a and the pattern ¡H*+L, the speaker would leave the door open to either an affirmative or negative response on the part of the interlocutor, while expressing his/her obliging disposition. It could be paraphrased as something like “You are free to accept or reject my proposal, but I want to let you know that I would like you to accept”. The use of the ¡H*+L L% pattern as an invitation/offer agrees with Armstrong (in press). In her study about polar questions in Puerto Rican Spanish, she finds that H+L* L% encodes positive epistemic bias but is also felicitous in the context of an offer/invitation. According to the author, by using this intonational pattern, the speaker can express his/her desire for the hearer to accept the invitation. This would not be problematic with the fact that H+L* L% in Puerto Rican Spanish encodes positive bias (as it is also the case for ¡H*+L L% in Logudorese Sardinian), because the definition of epistemic bias in Sudo (2013) includes both beliefs and desires (Armstrong, in press).

The preferred pattern in the epistemic conditions is the ¡H*+L L% intonation produced on a polar question with constituent fronting (e.g., Custu lìberu, dae munitzìpiu, battidu mi l’as? ‘Regarding the book from the city council, have you brought it to me?’). Another possible strategy, especially in the negative epistemic condition, is the ¡H*+L L% pattern, this time realized on a polar question with neutral word order (e.g., Mi l’as battidu, su lìberu? ‘Have you brought it to me, the book?’). In the neutral condition, we interpreted the ¡H*+L L% pattern as a carrier of “bias”. Now we add the notion of “positive”, that is, leading to a positive response, conveyed by the constituent fronting. This idea is reinforced by the fact that in negative epistemic contexts, we find a slight decrease in constituent fronting (17/23 vs. 15/26 which means 73% vs. 57% of the total data).This is correlated with an increase in the number of questions with neutral word order (Figure 7).

In the evidential condition two different patterns are found depending on the polarity of the bias condition, that is, whether the question conveys the speaker’s bias towards a positive or negative answer. When the speaker’s bias is towards an affirmative response, as expected, the preferred strategy is the ¡H*+L L% in combination with constituent fronting (e.g., Ma in pintzione ses andada? ‘Have you retired?’). However, when the speaker’s bias leans towards a negative response, we find a negative polar question (Ma no nde bendides pius, de frùtture e de birdura, como? ‘You will not sell them anymore, fruit and vegetables?’) with either the ¡H*+L L% (Figure 13) or the ¡H+L* L% intonation. The appreciable number of questions with ¡H+L* L% intonation (5 out of 19 polar questions) produced in the negative evidential condition could be explained, as kindly suggested by a reviewer, by the fact that all the contexts we used to elicit polar questions in this condition were contexts in which the speakers might be surprised (the speaker was always getting bad news; someone might be fired or a shop was about to close). One could hypothesize, then, that this contour can also be used to convey surprise in this language variety. In any case, this is only a conjecture and perceptual evidence would be needed to prove that this is the case.

Table 3 offers a summary of the lexico-syntactic and intonational linguistic markers found in our data as well as their associated meaning. Our data allow us to disentangle the specific contribution of the particle a to polar questions. Up to now this particle had been characterized as an encoder of positive polarity (with no clear difference relative to focus fronting; Remberger, 2010), as a particle heading requests, invitations and offers (Jones, 1993) or as having the function of assigning “focal value to the content in its scope” (Floricic, 2009). Our own interpretation, however, is that the particle a is a mitigator or politeness marker that mitigates the request for information in neutral polar questions or gives the hearer the possibility of accepting or refusing the invitation in offers. Our proposal is based on two observations derived from our data. The first one is that this particle does not appear in pragmatic contexts other than neutral. This indicates that the meaning of this particle cannot be associated with speaker bias, otherwise it might be felicitous in a situation where the speaker has epistemic or evidential bias leaning towards an affirmative response. The second observation is that it is not the particle alone which is responsible for the conveyance of offers or invitations, but rather the combination of the particle with the ¡H*+L L% intonation.

With respect to constituent fronting, our results agree with the proposal in Remberger (2010). Thus, it encodes “positive focus” in terms of Remberger or that the speaker expects it to be true that p (Remberger, 2010, p. 571). This explains why constituent fronting is very residual (only two cases) in neutral contexts or contexts in which the speaker has no bias. More evidence in favor of “positive bias” as the meaning associated with constituent fronting is provided by the fact that this strategy is incompatible with negation (see Floricic, 2009; Remberger, 2010).

Whereas constituent fronting is related to positive bias, negation and neutral word order (no fronted constituent) seem to be clearly related to negative bias, as we observe a significant increase in the use of this strategy especially in the negative evidential context.

As for intonation, the present data allow us to refine the analysis presented in Vanrell, Ballone, et al. (2014, in press). In those studies, the ¡H+L* L% pattern was related to unbiased polar questions (called also information-seeking yes-no questions in Vanrell et al., in press), whereas the ¡H*+L L% pattern was related to biased polar questions (confirmation or echo polar questions in Vanrell et al., in press). We propose now that the ¡H+L* L% pattern is a marker of “lack of bias”. Given that this pattern could also be produced in negative evidential situations, we hypothesize that it can also be used to convey surprise. Further research is required to provide evidence in favor of this hypothesis. The ¡H*+L L% pattern remains associated with the speaker’s bias either towards an affirmative (in conjunction with constituent fronting) or a negative (with negation) response. In offers or invitations, the particle a acts as a politeness marker, avoiding imposition, whereas the ¡H*+L L% contour would be related to a certain predisposition to acceptance.

The first stated objective of this paper was to offer a more refined analysis than what is presented in Vanrell, Ballone, et al., (2014, in press) for the ¡H+L* L% and the ¡H*+L L% associated with polar questions. This objective has been achieved by proposing the interpretation of “lack of bias” for the ¡H+L* L% intonation, compared to ¡H*+L L%, which is associated with the expression of speaker bias. ¡H+L* L% could also be related to surprise, specifically, to negative surprise caused by getting bad news. With respect to the second goal, which was to test the assertion that there exists a one-to-one correspondence between lexico-syntactic and intonational structures, we have provided evidence against the idea that the intonational form in Sardinian polar questions is determined by the lexico-syntactic structure. By contrast, what we see is that the same lexico-syntactic structure can bear different contours and that it is the combination of a particular lexico-syntactic form with a contour that determines the meaning of the utterance. This has typological implications in that it demonstrates that (a) languages can employ a combination of intonation and other mechanisms (such as question particles or changes in word order) to mark polar questions and (b) the use of these mechanisms can be with the expression of not only sentence modality but also a speaker bias towards an affirmative or negative response. Finally, we have provided a tentative pragmatic interpretation of the multiple mechanisms existing in Sardinian to construct polar questions. Further research could be carried out to determine whether the pragmatic analysis provided here also works at the perceptual level.

This article has described the intonation patterns of polar questions in Sardinian by considering also its interaction with syntactic structure and the appearance of syntactic particles. We claim that the joint analysis of all these components is needed in order to obtain an integrative view of how prosodic patterns work together with other grammatical components in natural languages. The results demonstrate that the different lexico-syntactic and intonational choices correspond to the expression of speaker bias in yes-no questions. Specifically, we have demonstrated that regarding the lexico-syntactic markers, the particle a acts as a mitigator or politeness marker, whereas constituent fronting and negation correspond to positive and negative bias respectively. With respect to intonation, two different patterns are distinguished, the ¡H*+L L% pattern and the ¡H+L* L% pattern, which represent 79 and 22 out of 112 cases respectively. The ¡H+L* L% pattern is interpreted as marking “lack of bias”. By contrast, the ¡H*+L L% pattern is restricted to situations in which the speaker is expressing his/her bias towards an affirmative or negative response. Interestingly, no specific marker was found that distinguishes between epistemic and evidential bias.