1 Introduction

This paper has two inter-related foci, one specific to Hindi-UrduFootnote 1 and the other more general. Its empirical focus is a particular lexical item in Hindi-Urdu that we refer to as polar kya:. We identify various syntactic restrictions on its occurrence and provide a descriptively adequate account of those restrictions. Its theoretical contribution is to leverage the account of polar kya: to draw a distinction between two types of interrogative particles that have often been grouped together. One type of interrogative particle is the one typically referred to as a Q-morpheme. We take this to be the overt realization of C[+Q]. The other we call a polar question particle (PQP), which occurs only in a subset of clause-types marked C[+Q]. The first class is well-established in the literature, with Japanese -ka and -no as prototypical examples. The second class is exemplified by Hindi-Urdu polar kya:, the new kid on the block.

We begin by introducing the signature properties of the Hindi-Urdu polar question particle in Sect. 2: its restriction to polar questions, its flexible syntactic positioning, and its selectiveness in appearing inside embedded polar questions. In Sect. 3, we present diagnostics distinguishing polar question particles from a clause-typing Q-morpheme. We analyze the Hindi-Urdu PQP kya: as having a presupposition that targets polar questions exclusively and we locate it high in the left periphery, at a structure with the prosodic profile of matrix polar questions. We also account for its clause-internal distribution. In Sect. 4, we consider the pragmatic contribution of kya:. We present diagnostics to establish that it partitions the clause into a segment that is not open to challenge and a segment that is unspecified in this respect and may be challenged. The precise distinctions that arise from this partition are shown to follow from independently motivated constraints that are operative in the grammar of Hindi-Urdu. Section 5 deals with two issues that arise when we consider polar kya: in connection with disjunction. One bears on the presupposition we ascribe to it, the other is an unexpected restriction against final kya:. We conclude in Sect. 6 by noting two domains of inquiry that our analysis of polar question particles opens up. The overall message we try to highlight is that Hindi-Urdu polar kya: belongs to a class of items that has not so far been given formal recognition but for which there is cross-linguistic evidence in the literature.

2 Properties of the polar particle kya:

In this section we establish that the particle kya: is restricted to polar questions, that it can occur almost anywhere within a clause, that it can appear in matrix as well as embedded clauses, albeit with some restrictions.

2.1 Hindi polar questions and kya:

To an initial approximation, polar questions, also known as Yes/No questions, have rising intonation on the verbal complex in Hindi-Urdu, whereas declaratives generally have falling intonation.Footnote 2 Butt et al. (2017) and Biezma et al. (2017) associate rising intonation with L/H-H% and falling intonation with L-L% in (1).

  1. (1)
    figure a

Unlike English, they do not involve inversion of the finite verb. The characteristic prosody noted above is, however, obligatory for a Y/N question interpretation. Y/N questions optionally co-occur with the wh-word kya:. It should be noted that the presence of the polar particle kya: does not make the characteristic prosody optional. Note also that despite the presence of this prosody, Hindi-Urdu polar questions are neutral questions unlike English declarative questions (Bartels 1997; Gunlogson 2003, see also Jeong 2018; Farkas and Roelofsen 2017, and Westera 2018).

  1. (2)
    figure b

In (2), kya: is not the argument of any predicate. But kya: can also function as an argument of a predicate with the meaning ‘what.’

  1. (3)
    figure c

To distinguish these two cases, we dub the athematic kya: in (2) the polar question particle kya: and gloss it as ‘PQP,’ short for ‘Polar Question Particle.’ The kya: in (3), we call thematic kya:. Butt et al. (2017) note that the polar particle kya: has a flat intonation while the thematic kya: has a H* pitch accent, which accent also appears more generally on wh-phrases in Hindi-Urdu.Footnote 3

Polar kya: does not appear in constituent questions.

  1. (4)
    figure d

Polar kya: can appear in alternative questions; we will argue that its appearance there follows from its appearance in Y/N questions.

  1. (5)
    figure e

2.2 The basic distribution of polar kya:

The most unmarked location for polar kya: is the clause-initial position. But it can appear in almost any other position. It can be clause-medial or clause-final.Footnote 4

  1. (6)
    figure f

In an almost mirror image pattern, thematic kya: is natural in the immediately pre-verbal position but odd/marked elsewhere.Footnote 5

  1. (7)
    figure g

All else being the same, one might expect that the distribution of polar kya: in embedded clauses would simply track the distribution of embedded Y/N questions. However, this expectation is not borne out. To a first approximation, polar kya: can only appear in complements of rogative predicates, predicates that take only interrogative complements, but not in complements of responsive predicates, predicates that take interrogative as well as declarative complements. Note that the Hindi-Urdu complementizer ki tracks finiteness but is otherwise compatible with declarative, interrogative and subjunctive clauses.

  1. (8)
    figure h
  1. (9)
    figure i

Note that in (9b), the predicate that takes the embedded question as a complement is ja:n ‘know’, which is a responsive predicate but in combination with the attitude predicate ca:h ‘want’ functions like a rogative.

2.3 Embedded polar kya: and embedded inversion in English

We confirmed the unavailability of polar kya: in complements of the responsive predicate ja:n ‘know’ by searching the Corpus Of Spoken Hindi (COSH) using the COSH Conc [Software].Footnote 6 There was no shortage of embedded kya: questions in the corpus but we did not find cases like (8). We did find cases of polar kya: under responsives but only under specific conditions. In each of the following examples, the responsive combines with another operator.

  1. (10)
    figure j

The responsive versus rogative distinction that we see in the distribution of embedded polar kya: has been noted by McCloskey (2006) to play a role in the distribution of embedded inversion in English.Footnote 7

  1. (11)
    figure k

As indicated in the schematic version of these examples, McCloskey takes the possibility of embedded inversion as indicating the presence of additional CP-structure but he goes on to note that the distribution of embedded inversion cannot be reduced to the choice of embedding predicate. As the following example shows, the responsive predicate know can combine with embedded inversion if it is part of a larger structure want to know.

  1. (12)

    Everybody wants to know [did I succeed in buying chocolate for Winifred].

A question that arises about such examples is whether the complement clause is a quotation. McCloskey gives detailed arguments against this possibility from pronominal binding and sequence of tense. Here we note that if the complement clause in (12) was a quote, I would be bound by everybody but this particular sentence is noted by McCloskey to be a a naturally occurring sentence where I refers to the speaker in the context.

McCloskey also notes that questioning or negating the responsive improves the acceptability of embedded inversion.

  1. (13)
    1. a.

      *I remember was Henry a communist.

    2. b.

      Do you remember was Henry a communist?

    3. c.

      ?I don’t remember was Henry a Communist.

It is striking that in Hindi-Urdu some of the same features modulate the acceptability of embedded polar kya: with responsives, as we have already seen in (9b) and (10). In fact the parallelism extends to question complements of nouns. Compare (14a) from COSH with (14b) from Gunlogson (2003).

  1. (14)
    figure l

Further examples could be constructed to show the parallel between English embedded inversion and Hindi-Urdu embedded PQPs but we believe that we have established the pattern sufficiently. The distribution of the Hindi PQP kya:, then, is not as arbitrary as it might seem at first glance. When seen in the broader perspective of what Dayal and Grimshaw (2009) label quasi-subordination, it appears quite systematic. It appears in matrix polar questions and in quasi-subordinated embedded polar questions. We started with the responsive/rogative distinction but it has turned out not to be what underlies the distribution of embedded inversion in English and embedded kya: in Hindi-Urdu. We have already seen examples of modified responsives that can take kya: complements and Dayal (2019) notes that there is a class of rogative predicates that do not allow embedded inversion in English or embedded kya: in Hindi-Urdu.

  1. (15)
    figure m

3 The syntactic and semantic contribution of polar particle kya:

In this section we work towards an account of polar particle kya:, focusing on its distributional properties. We first discuss the possibility that it functions as a clause-typing particle. We then offer an alternative in which it is a particle that is situated higher in the left periphery than CP and which selects for singleton propositional sets. We also discuss how kya: can appear in different positions in the clause.

3.1 What polar particle kya: is not

A plausible first pass at analyzing polar particle kya: is to treat it as a Q-morpheme, that is optionally overt. Indeed, this is how Cheng (1991:21) characterizes its role. There is, however, good reason to doubt that kya: is a marker of the clause type interrogative. Consider the contrast in embedding possibilities that we have been studying.

  1. (16)
    figure n

We have seen that the contrast above represents a more general pattern. Polar particle kya: does not seem to embed under responsive predicates but it can be embedded under rogative predicates. We saw in Sect. 2 that matrix polar questions do not differ syntactically from their declarative counterparts. Their status as interrogatives is signaled by a rising vs. a falling intonation. However, the situation is different in embedded positions, where prosody cannot play the same role. (17) shows that responsive predicates cannot take such a complement under an indirect question interpretation.

  1. (17)
    figure o

We note that in order to get an indirect question interpretation, the embedded clause must be a polar alternative question, with an overt disjunction plus negation ya: nahĩ: ‘or not’. Interestingly, we cannot add kya: in this case, as shown in (18b):

  1. (18)
    figure p

The contrast between responsive and rogative predicates with respect to embeddability of kya: is replicated in Mandarin, with respect to the particle ma, which Cheng treats in the same terms as Hindi kya:.Footnote 8

  1. (19)
    figure q

The facts discussed here establish that Hindi-Urdu kya:, and Chinese ma by extension, is not the yes/no operator. If it were, it would be able to occur optionally or obligatorily in all embedded positions. It is also not a straightforward clause-typing particle, marking an interrogative clause. If it were, one would again expect it to occur in all cases of embedding. One could even argue that such a particle might be obligatory when embedded under responsive predicates where there may be more functional pressure to distinguish between declarative and interrogative structures.

The crucial point for us is to distinguish Hindi-Urdu kya: and Chinese ma from straightforward Question Particles like Japanese -ka/-no, which are not subject to the restrictions found with kya:. The Japanese Q-morpheme, then, exemplifies the kind of clause-typing that was proposed by Cheng (1991).

  1. (20)
    figure r

One analysis that is compatible with the restricted distribution of polar particle kya: and may still qualify as a form of clause typing is to take kya: to occur in a projection above CP, let’s call it ForceP. We wish to emphasize though that this would be different from clause-typing of the Japanese kind which targets all questions. In fact, this account of polar kya: mirrors the analysis proposed by McCloskey (2006) to explain the parallel facts about inversion in English discussed in Sect. 2.3. McCloskey entertains the possibility that some predicates can only take regular CP complements that denote questions, while others may also take ForceP complements that denote the question speech actFootnote 9:

  1. (21)
    1. a.

      (wonder): [[C []]]

    2. b.

      (know): [C []]

Our claim, then, is that kya: is only acceptable in the complements of predicates that can take ForceP. These are canonically the set of rogative predicates. However, as we have seen, there are some cases where kya: can occur in the complement of responsive predicates if those predicates are negated or questioned. We have also seen the case of a rogative predicate that does not take embedded kya:. We set aside the issue of why selection of a complement should be a fluid matter, and focus here on the fact that the fluidity is entirely systematic and in keeping with the proposal to treat kya: as occurring in a projection above CP, which we are taking to be ForceP.Footnote 10

3.2 What polar kya: is

The issue we will now address is the fact that polar particle kya: is acceptable in Y/N questions but not in wh-questions. In addressing this, we will assume that Y/N questions differ from wh-questions in denoting singleton sets of propositions:

  1. (22)
    figure s

The view that wh-questions denote a set with more than one member is standard. It rests on the view that wh-phrases must range over a plural domain in order for the conditions for proper questioning to be satisfied. That is, a question like (22a) carries an existential presupposition that someone left and an uncertainty presupposition about who that person may be. An insight going back to Bolinger (1978), and recently revived in the literature, is that Y/N questions are fundamentally different in privileging the nucleus proposition. A yes answer or a no answer can then be treated as anchored to this proposition. See Gawron (2001), van Rooij and Šafářová (2003), Biezma (2009), Farkas and Bruce (2010), Biezma and Rawlins (2012), Krifka (2013) and Roelofsen and Farkas (2015). In line with the spirit of these works, we propose that the basic denotation of (22b) would be the singleton set {John left}.

We now show how this independently motivated distinction can help explain the distribution of polar kya:, drawing on the account of similar restrictions discussed by Xu (2012, 2017) in connection with the Mandarin particle nandao. We define kya: as an expression that encodes a presupposition that its complement is a singleton set:Footnote 11

  1. (23)

    \(\mbox{$[\![ \text{\textit{kya:}} ]\!]^{}$} = \lambda Q_{(st)t}: \exists p \in Q [\forall q [q \in Q \to q = p]]. Q\)

Since kya: is defined on a set of propositions, it rules out kya: with declarative statements. It further restricts kya: to a subset of questions, namely those with just one proposition. Wh-questions, as we have just discussed, do not pass the requirement of singularity.Footnote 12 Interestingly, neither do alternative questions. This may appear to be a potential problem for our account as we have seen that kya: is fully acceptable in alternative questions like (5). We will return to this in Sect. 5, where we will see that appearances notwithstanding, the pattern falls well within the terms of our account.

Our proposal relies on a type distinction between declaratives and interrogatives. We have further assumed that polar questions are distinguished by being singleton sets from other types of questions. A reviewer points out that it is possible to distinguish polar questions from declaratives as well as other kinds of interrogatives semantically without making these assumptions. This can be done by replacing the singleton requirement of kya: by a requirement that kya: presupposes that the clause it combines with highlights a single proposition. The notion of highlighting is as in Roelofsen and Farkas (2015). This would be the way forward in theories which do not make a type distinction between declaratives and interrogatives (see Roelofsen 2019 and Uegaki 2019). In fact, Xu (2017), that we mentioned earlier, does precisely this for Mandarin nandao.

3.3 Intonation and Hindi-Urdu Yes/No Questions

At this point, we would like to clarify our assumptions about the connection between our syntactic proposal and the intonation associated with Hindi-Urdu Yes/No Questions. We do this by introducing a problem in the syntax and semantics of Hindi-Urdu that is independent of the issue of polar question particles. Consider the paradigm in (24).

  1. (24)
    figure t

The examples in (24a, c) tell us that Hindi-Urdu has the same pattern of embedding questions as English. For this reason we allow Hindi-Urdu responsive predicates to take CP complements headed by C[+Q], where we take \(\mbox{$[\![ \text{C[+Q]} ]\!]^{}$}\) to denote λqλp[p = q]] as proposed, for example, in von Stechow (1996). The only point of cross-linguistic variation is the indirect question interpretation of the polar question in (24b). We believe that in English the presence of the complementizer if/whether allows for an indirect question interpretation while Hindi-Urdu requires matrix clause intonation for this purpose. Footnote 13 This makes embedding polar questions, in effect, a root phenomenon in Hindi-Urdu since the indirect question interpretation piggy-backs on rising intonation, which may well be a root phenomenon cross-linguistically.Footnote 14

Here is a prediction that follows from our account. Consider the following sentence.

  1. (25)
    figure v

In principle, the complement in this sentence has two possible parses, as [C…] and as [↑ [C…] . But the first option (=CP) is ruled out analogously to (24b). The second option (=ForceP) saves the sentence but it brings with it a short prosodic break after the matrix clause and the prosody of matrix polar questions on the complement. And, not surprisingly, it becomes possible to have kya: in the embedded clause: ki kya: dharti: gol hai.

The crucial role of intonation in licensing monoclausal yes/no questions raises issues that are independently interesting. We refer the reader to Dayal (2019) for further discussion. For present purposes the following points are important. Rising intonation of the kind seen in matrix Yes/No questions is associated with ForceP[+Q], in embedded contexts rising intonation shows the hallmarks of embedded root phenomena and, to the extent that we only see kya: in the presence of rising intonation, we consider PQPs to also be an embedded root phenomenon.Footnote 15 Note that a ForceP[+Q] associated with a Yes/No question may or may not have a kya:. But such a ForceP is always associated with rising intonation.Footnote 16

3.4 Deriving non-initial kya:

We now come to the third part of the distributional puzzle, which has to do with the position of the polar question particle kya: within the clause it occurs in. The various possibilities for the occurrence of kya: can be derived by assuming the base structure in (26) followed by movement to the left of kya: as illustrated in (27).

  1. (26)

    [kya: [C[+Q] [anu-ne uma-ko kita:b di:]]]

  1. (27)
    figure w

Some support for the movement proposal comes from the fact that kya: is difficult after weak indefinites like kuch ‘something’ and idiomatic expressions, which are elements whose movement leads to deviance.Footnote 17

  1. (28)
    figure x

According to our proposal, the appearance of the polar question particle kya: in an immediately pre-verbal position in (28) indicates that the pre-kya: material has moved over kya:. In (28), we have direct objects that are resistant to movement, as shown by the following contrasts.

  1. (29)
    figure y
  1. (30)
    figure z

Since the direct objects in (28) are resistant to movement, it follows that the variants of (28a, b) where the polar particle kya: is immediately pre-verbal are deviant. That the deviance of immediately pre-verbal polar particle kya: is related to the movement potential of the direct object is shown below. Here we have an object that can move freely and we find that kya: can appear in an immediately pre-verbal position without deviance.

  1. (31)
    figure aa

Following a similar logic as for the above examples, clause-final kya: is derived by scrambling of the whole finite clause to the left of kya:.

  1. (32)
    figure ab

This accounts for the attested word order variations observed with respect to kya:.

4 The pragmatic contribution of polar particle kya:

In this section we consider the pragmatic contribution of kya:. Starting with non-initial kya: we present diagnostics to establish that kya: partitions the clause into a segment that is not open to challenge and a segment that is unspecified in this respect and may be challenged. The precise distinctions that arise from this partition are determined by independently motivated constraints that are operative in the grammar of Hindi-Urdu.

4.1 kya: Induced partitions

When kya: is clause-initial (or absent), an alternative question can be formed on any element in the clause.

  1. (33)
    figure ac

When kya: is not clause-initial, alternative questions can only be formed on phrases that follow but not those that precede it—providing alternatives to them using the above syntactic frame is unacceptable. We illustrate this in (34), where kya: follows the subject and the indirect object but precedes the temporal adverb and the direct object.

  1. (34)
    figure ad

The partition contrasts found with gapping are replicated in a paradigm where we consider possible negative responses to a Yes/No question with kya:. With initial kya:, any phrase can be targeted for correction. This is shown in (35).

  1. (35)
    figure ae

But when kya: is clause-medial, the constituents which precede it cannot be corrected while the post-kya: elements can be.

  1. (36)
    figure af

The pattern we have noted so far holds if the pre-kya: material is read with neutral intonation. But prosody can make a difference. Biezma et al. (2017) note that when the material to the left of kya: is stressed, it is in fact possible to correct/challenge it as in (37).

  1. (37)
    figure ag

We agree with the judgement that they report in (37) but we do not think that prosody is sufficient in general to make just any pre-kya: element contrastable. Consider (38).

  1. (38)
    figure ah

Prosodic prominence on the subject in (38) does not make it open to challenge. The focus marking on Ram feels unmoored; it does not help in identifying the contrasted element. Footnote 18

Comparing (37) and (38), we see that there is an adjacency effect for focus on the left: stress on the immediately pre-kya:-XP makes it contrastable but this is not an option for other pre-kya:-XPs. Syed and Dash (2017) claim that there is a similar adjacency effect on the right i.e. only the immediately post-kya: XP can be contrasted. But the data in (33b-d) and (34d), which are robust and widely-accepted (see Biezma et al. 2017), show that adjacency is not required on the right of kya:. To sum up, we claim that all post-kya: material and only the immediately pre-kya: XP can be contrasted. Before we turn to an explanation of this generalization in Sect. 4.3, we briefly explore the theoretical underpinnings of our analysis.

4.2 Deriving the partitions

We have argued that pre-kya: material gets to its surface position via movement. Moreover we have shown that being in a pre-kya: position is associated with certain interpretive restrictions. What motivates these movements? And where do these interpretive restrictions stem from?

It is attractive to connect the answers to both these questions to properties of kya:. Perhaps kya: has syntactic features which look for constituents that bear a matching feature whose semantic reflex is that the constituent is interpreted as backgrounded and hence not being open to challenge. kya: would attract any constituent in its local domain with such a feature and move it to its specifier. This would require us to treat kya: as a head in ForceP. But we believe that this proposal faces a number of challenges.

The first challenge is that what precedes kya: is not necessarily a single constituent. In principle, there can be any number of constituents preceding kya:. This makes a feature based account difficult to maintain. A deeper problem is that the pre-kya: domain is not homogenous. As we have discussed in Sect. 4.1, immediately pre-kya: material can go either way with respect to the challengability diagnostic depending upon the presence of focus.

For these reasons, we are partial to a different implementation, one that does not give kya: all that much to do. In this conception, the movement of phrases in Hindi-Urdu takes place for a variety of reasons—this is a language with scrambling, which we know to be a non-uniform process with a range of motivations. The fronting of phrases that we see with kya: is then not motivated by kya:. What kya: does, however, is reveal that the movement is to a location past ForceP. The interpretive effects are associated with material being in a location past ForceP; they are not encoded in the meaning/lexical entry of kya:.

4.3 Deriving contrastability through focus

According to our syntactic proposal, kya: is in ForceP and pre-kya: material gets there via movement to a position higher than kya:. We add to our analysis that kya: demarcates the domain that can be focused, which is minimally its c-command domain. This means that in the schema in (39a), YP and ZP can be focused. Deferring discussion of XPfor the moment, we can also say that XPcannot be focused.

  1. (39)
    figure ai

Recall that we have taken the ordinary semantic value of polar questions to be a singleton propositional set. The response particle yes is anaphoric to the unique member of this set and asserts it. The response particle no is also anaphoric to this proposition but denies it. What we need to add is a characterization of how permissible corrections arise.

We follow the literature in taking prosodically stressed elements to create a set of alternatives. The focus semantic values shown in (39b) and (39c) draw on alternatives to YP and ZP respectively. A plausible continuation of a no answer draws on such sets. In this we are essentially following the analysis of Turkish polar questions with mı in Atlamaz (2015). In making reference to the focus semantic value of questions, our analysis also shares properties with the proposal in Biezma et al. (2017) which is couched in terms of Questions under Discussion.

Turning now to XP, we note that Hindi-Urdu has several particles that associate to the immediate left: hi: ‘only’ (Bajaj 2016), bhi: ‘also’ (Dayal 1995; Lahiri 1998), nahiĩ ‘not’ (Kumar 2003) among others. Given this broader perspective on the grammar of Hindi-Urdu, we conclude that the paradigm regarding kya: is not unexpected. Given appropriate prosody, the focus semantic value of (37) will include alternatives generated by the immediately pre-kya: XP: {Anu gave Uma a present, Ram gave Uma a present, Vina gave Uma a present, …}.

Allowing the immediately pre-kya: constituent to be open to challenge has interesting implications for another structure that our account makes available. Consider final kya:.

  1. (40)
    figure aj

An important difference between the fronting seen here and the cases discussed earlier in (39) is that the fronted constituent is a clause and is therefore able to be have internal prosodic structure (for an overview of theories relating prosody to information structure see Büring (2017); for studies specifically looking at Hindi-Urdu see Féry et al. 2016, Patil et al. 2008, and Genzel and Kügler 2010). With appropriate stress on Uma as in (41), for example, the focus semantic value can shift even though Uma neither follows kya: nor is it left-adjacent.

  1. (41)
    figure ak

This focus semantic value allows for corrections and alternatives to Uma.

We have ended up with near synonymy between clause-initial and clause-final kya:, which can both allow alternatives and corrections on any focused constituent in their focus domain. As we will see in the next section, however, the parallelism breaks down when polar kya: interacts with disjunction.

5 Polar kya: and alternative questions

There are two issues that arise when we consider polar kya: in connection with disjunction. One bears on the singleton-set requirement that we have claimed for it, the other is an unexpected restriction against final kya:.

5.1 The singleton set requirement and disjunction

In Sect. 3, we alluded to Bolinger in connection with the difference between polar questions as singleton propositional sets and wh-questions as non-singletons. Let us consider now polar questions with disjunction.Footnote 19

  1. (42)
    figure am

The distinction between Y/N questions and Alternative Questions is not always easy to make but they can be identified on the basis of prosody. In English, Y/N questions have a rising intonation while Alternative Questions have pitch accents on the two alternatives and a final fall. In Hindi too, there is a similar prosodic difference, which we have indicated in (42b) using square brackets. Additionally, Hindi has two lexical items for disjunction, ya: and ki. While ya: can occur in both types of questions, ki can only occur in alternative questions, not in Y/N questions or declarative sentences. We use capital OR to indicate those instances of disjunction where only an alternative reading is available. (42b) shows that kya: is compatible with alternative questions.

This may be a good place to clarify our position on how the prosodic features of alternative questions relate to structure. Bartels (1997) identifies three signature properties, a pitch accent on each disjunct, a prosodic break between the disjuncts, and a final fall. Pruitt and Roelofsen (2013) establish that pitch accents and the prosodic break are more significant than the final fall in identifying alternative questions. From our perspective, the placement of pitch accents and prosodic break is a clause-internal phenomenon that we expect to be always in evidence, in matrix alternative questions as well as alternative questions embedded under rogative or responsive predicates. We have argued that the matrix intonation contour enters the derivation at ForceP, which is also the position at which we see PQP kya: in Hindi-Urdu and embedded inversion in English. We would venture to say that this is also where the final fall that marks the closure of proffered options in alternative questions is located. Since this final fall is indistinguishable from lack of matrix intonation in complements of predicates that do not embed ForceP, the final fall on its own does not help us separate out alternative questions that project up to CP from those that project up to ForceP. Interestingly, Ciardelli et al. (2019) also note cases of alternative questions that have a final rise. Such questions, however, have the behavior of root phenomena. This adds further support to our claim that matrix question intonation is located in ForceP, even in alternative questions.

With respect to answerhood conditions, Alternative Questions typically expect a positive response to exactly one of the proffered alternatives but the choice between them is open. This is taken as evidence that each alternative is included in the question denotation.Footnote 20 If alternative questions denote multi-membered sets, the Hindi-Urdu PQP kya:, as we have defined it, should be incompatible with them. So the fact that they are in fact compatible calls for an explanation. Our solution rests on the view that it is possible to analyze an alternative question with kya: as (44) instead of (43), optionally followed by ellipsis of material in either CP1 or CP2. The account of Hindi-Urdu alternative questions that we develop below is directly inspired by Han and Romero (2004).

  1. (43)
    figure an
  1. (44)
    figure ao

For completeness, we add the structure with kya: and disjunction inside a polar question in (45) below but in what follows we will set aside Yes/No interpretations.

  1. (45)
    figure ap

In support of our proposal, we note that alternative questions can be conveyed by what looks like an explicit disjunction of two Yes/No questions.Footnote 21 In Hindi-Urdu also, we find that two kya: questions can be disjoined to yield an alternative question.

  1. (46)
    figure aq

This disjunction of two kya: questions seems to have the same meaning as the version with just one initial kya:. In fact, as far as we can tell, all the following four variants are acceptable and can be used to convey alternative question readings, of course with the appropriate prosody.

  1. (47)
    figure ar

We can interpret all these alternatives with our assumption concerning polar questions and kya: and an assumption about how to interpret disjoined polar questions. In our analysis, p and q individually combine with C[+Q] to form polar questions. By the semantics we have assumed, this means that each polar question denotes a singleton set, {p} and {q} respectively. kya:, if present, only applies to the singleton set corresponding to one of the polar questions and as a result its requirement is met. The crucial component for us is that when we disjoin two Y/N questions, we end up with a multi-membered set {p,q}. (Also see Alonso-Ovalle (2006) and Krifka (2015) among others.Footnote 22)

5.2 Final kya: and disjunction

In Sect. 4.3 we saw that clause-initial kya: and clause-final kya: allow the same set of answers. One might therefore expect them to display similar behavior across the board. The two come apart rather spectacularly, however, in the context of disjunction. As we have seen, initial kya: is compatible with a disjunction of finite clauses. But final kya: is not.

  1. (48)
    figure as

Let us remind ourselves of what we take to be the syntactic structure of final kya:. According to our analysis, clause final kya: is derived by movement of the full clause to the left of kya:.

  1. (49)
    figure at

We want to understand the ungrammaticality of final kya: with disjunction of finite clauses, which we can schematically refer to as ‘p or q kya:’. The following parses are in principle available.

  1. (50)
    figure au

Limiting ourselves to alternative questions, we first consider the parse in (50a).

  1. (51)
    figure av

For the needs of kya: to be met in an alternative question, kya: must scope under disjunction but in this structure kya: scopes over it. We have now shown why (50a) is unavailable. What about the structure in (50b) (‘[p or [q kya:]]’)? where kya: is attached to the second disjunct i.e. the disjunction takes scope over ForceP.Footnote 23

  1. (52)
    figure aw

This derivation is well-formed and it predicts that the structure should have an alternative question interpretation, which is not in fact available. Why is such an interpretation unavailable? A structure very similar to it is, in fact, what we have argued underlies alternative questions with initial kya:, namely ‘[[kya: p] or q]’.

While we do not have an explanation for the unavailability of the ‘[p or [q kya:]]’ structure, we believe that restrictions on the disjunction of structures with internal topicalization play a crucial role. The significant difference between the unacceptable ‘[p or [q kya:]]’ and the acceptable ‘[[kya: p] or q]’) is that there is clausal topicalization over kya: in the first case. One might imagine that fronting over kya: and disjunction of polar questions would operate independently of each other. This turns out not to be the case. There seem to be stringent restrictions on what combinations of frontings are legitimate in disjoined polar questions. Unlike the well-formed disjunction of two kya:-initial polar questions in (53a), its medial kya: counterpart in (53b) is ungrammatical.

  1. (53)
    figure ax

The restriction on ‘p or [q kya:]’ is, from this perspective, part of a larger restriction on the interaction of disjunction and topicalization within the disjuncts.

6 Extensions

We would like to end our discussion by considering two domains in which our account calls for follow-up work. One is specific to polar kya: and involves the extension of our account beyond information-seeking questions. The other is the cross-linguistic application of the distinction we have made betwen PQPs and Q-morphemes.

6.1 Polar kya: and types of polar questions

We have established that polar kya: is restricted to polar questions but we would like to see if there are restrictions within the class of polar questions that may give us further insight into its semantics and pragmatics. We find that polar kya: is perfectly acceptable, for example, in polar questions used as rhetorical and quiz questions.

  1. (54)
    figure ay

In the above examples, kya: can occur in all the positions one might expect, of course with appropriate prosody.

We would like to point out that kya: is unacceptable in rhetorical uses of wh-questions. We follow the approach of Rohde (2006) and Caponigro and Sprouse (2007) in taking rhetorical questions to have the same denotation as information seeking questions. In our terms, this amounts to singleton sets as denotations for polar questions and non-singleton sets for wh-questions. The rhetorical use surfaces in contexts where one proposition in the set is known by both participants to be the only true member of the set. Given this approach and the singleton set requirement of kya:, it follows that PQP kya: is not compatible with rhetorical wh questions (see also Dayal 2016: 283-285 for alternative views of rhetorical questions).

kya: is also acceptable in negative and/or biased questions.Footnote 24

  1. (55)
    figure az

In fact, we have so far only been able to definitively identify one type of polar question where polar kya: is not possible. These may be classified as incredulity questions. Footnote 25 Consider the following in a context where the addressee was supposed to have left town and the speaker is surprised to see him:

  1. (56)
    figure ba

To be completely upfront about this, it is unclear to us whether (56) is a polar question or an exclamation. It certainly has a prosody very similar to a polar question. If it is a polar question, then the singleton presupposition requirement of kya: would be satisfied and its unacceptability would have to be traced to a different aspect of its pragmatic profile. If, however, (56) is an exclamation and denotes a proposition rather than a set of propositions, the unacceptability of kya: here would follow from the presupposition we have posited. We leave the precise status of (56) for the future, noting only the significance of direct evidence—the speaker directly witnesses the presence of the interlocutor— in regulating the distribution of kya:.

This admittedly brief discussion of the pragmatics of Yes/No questions with polar kya: resonates with, but is not identical to, the more detailed investigations of this topic conducted in a series of papers by Biezma et al. (2017).

6.2 Interrogative question particles crosslinguistically

In the preceding sections we have presented a fairly detailed account of the polar question particle kya: in Hindi-Urdu. In doing so, we have uncovered a complex set of interwoven grammatical effects: the restriction of the polar question particle to polar questions, its sensitivity to quasi-subordination, its appearance in alternative questions, and the relationship between its position in the clause and information structure. One may well wonder if this is simply a quirky phenomenon restricted to the grammar of Hindi-Urdu. In this concluding section we would like to suggest that polar question particles are a robust cross-linguistic phenomenon whose full character is still to be understood. We believe Hindi-Urdu polar kya:, as the first of its kind to be analyzed, can help in this process of discovery.

We take the following to be a necessary criterion for determining whether a particular lexical item in a language is a polar question particle: it should only occur in polar questions. There may well be other significant features associated with polar question particles, such as optionality, information structural effects, and/or selectivity in embedding. As indicated earlier, this rules out Japanese -ka/-no as a polar question particle. We have taken -ka/-no to be Q-morphemes, an overt realization of C[+Q], following a fairly well-established practice in the syntactic and semantic literature. There are other candidates, however, that may qualify as polar question particles. Syed and Dash (2017) have extended the analysis of Hindi-Urdu polar kya: in Bhatt and Dayal (2014) to similar particles in the closely related Indo-Aryan languages, Odia and Bangla. And we have already alluded to the possibility that Mandarin -ma may belong with polar kya: as a PQP. Mandarin nandao, is another candidate for PQP, though it also obligatorily introduces bias (Xu 2017). Other candidates are Turkish -mi (Aygen 2011; Kamali and Büring 2011; Göksel and Kerslake 2004; Atlamaz 2015; Özyıldız 2018), Italian che (Nicoletta Loccioni, Paolo Crisma, Giuseppe Longobardi p.c.) and Slovenian kaj (Adrian Stegovec p.c.). The latter two are homophonous with the interrogative pronoun that means what, similar to the situation in Hindi-Urdu, Bangla and Odia.

We would like to end this discussion with a broader question: why do languages have Polar Question Particles? What would be lost if they did not? Do they add any expressive power? Our account has shown that the Hindi-Urdu kya: does not add any identifiable component of meaning. Its presence correlates with information structure effects but does not seem to cause those effects; rather it just makes them visible. To the extent that PQPs in other languages show a similar profile, we have evidence that natural language lexicalizes items that may not themselves contribute very much but their existence is justified in virtue of the mere fact that they hold up a mirror to other independently available processes.