3.1 Nakai questions

The particle nakai has one main fixed position in post-predicate position (Seiter Reference Seiter1980, Sperlich Reference Sperlich1997, Massam Reference Massam2003), as in (12).

1. (12) Kua kai nakai e Moka e apala.

   perf eat q1 erg.p Moka abs.c apple

   ‘Did Moka eat the apple?’(Massam Reference Massam2003: ex. (1))

It cannot occur in wh-questions, as in (13).

1. (13) *Ko hai nakai e tagata i kō?

   pred who q1 abs.c person gl.p dem

   ‘Who is that person there?’(Massam Reference Massam2003: ex. (6c))

The Q particle cannot occur in questions negated by either of the sentential negators nākai or ai (Massam Reference Massam2003), as illustrated in (14).

1. (14) *Nākai/ai kai nakai e Moka e apala?

   neg eat q1 erg.p Moka abs.c apple

   ‘Didn't Moka eat the apple?’(Massam Reference Massam2003: ex. (4c))

The fact that nakai cannot co-occur with ai suggests that the restriction is a structural one between nakai and NEG and is not due to the simple similarity in form between nakai and nākai. This is an important point that we will return to later, in Section 4.2.

When the discourse context of the yes–no questions is examined, we find that nakai forms neutral unmarked questions. In oral interviews, nakai questions tend to request information on new topics, as in (15), where the interviewer is changing the subject to ask a new, neutral question.

1. (15) Fā fenoga atu nakai a koe ke he motu ko Niue?

   hab journey dir2 q1 abs.p 2.sg gl.c loc.c island pred Niue

   ‘Do you usually travel to Niue?’(NMMIP5)

A second use for nakai is in a fixed tag, which can occur alone, or as a tag on an existing nākai question, as in (16). Q and P at the start of each turn mark questioner and participant, respectively. The translations provided are those of the translator, and they have not been tampered with, in order to preserve the translator's sense of the meaning of the original Niuean sentence. We often include the discourse context and answers, even though these make the examples long, as they provide clues as to the discourse uses of the various Q particles.

1. (16) Q: mitaki, fā fano nakai a koe ke he tau

   good usual go q1 abs.p 3.sg gl.c loc.c pl

   fakafetuiaga he tau Niue, e po.ke nākai?

   event gen.c pl Niue yes or neg

   Q: ‘Right, do you go to other Niuean events?’

   P: fano

   go

   P: ‘(I) go.’(Massam et al. Reference Massam, Starks, Ikiua, Moyse-Faurie and Sabel2011: ex. (9))

3.2 Kia questions

The other two Q particles are more flexible in their position. As mentioned earlier, kia can occur in post-predicate or in sentence-final position, as illustrated in (17)–(18).Footnote ¹²

1. (17) … Ko e vagahau Niue tūmau kia a koe?

   pred c language Niue always q2 abs.p 2.sg

   ‘Do you always speak in Niuean?’ (NOFIP5)

1. (18) Kai e Mele e apala kia?

   eat erg.p Mele abs.c apple q2

   ‘Did Mele eat the apple?’(Massam Reference Massam2003: ex. (2))

It should be noted that the latter position is relatively rare, in fact it has been found thus far only in our elicitation. Unlike nakai questions described earlier, kia can occur in wh-questions, as in (19).

1. (19) Ko hai kia ne manako a Moka ki ai?

   pred who q2 nfut like abs.p Moka loc.p resprn

   ‘Who does Moka like?’(Massam Reference Massam2003: ex. (6))

This would suggest that kia is a general question word (like Japanese ka) rather than a yes–no Q particle in the strict sense. Strictly yes–no particles are relatively rare in the languages of the world (Haspelmath et al. Reference Haspelmath, König, Oesterreicher and Raible2001), but appear to be also attested in at least one other Polynesian language, Samoan. Given the syntactic distribution, Otsuka (Reference Otsuka2006) argues that Samoan does not have a yes–no Q particle but rather a general Q marker.

A further characteristic of kia questions relates to negation. The form kia can occur in negated questions, with a NEG particle nākai or ai. In such questions, as noted above, the Q particle can appear in its usual post-verbal position, as in (20), or can appear pre-verbally after NEG, as in (21).

1. (20) Nākai kai kia e Mele e apala?

   neg eat q2 erg.p Mele abs.c apple

   ‘Didn't Mary eat the apple?’(Massam Reference Massam2003: ex. (4a))

1. (21) Ai kia mohe a Moka he mohega tote?

   neg q2 sleep abs.p Moke loc.c bed small

   ‘Didn't Moka sleep in the small bed?’(Massam Reference Massam2003: ex. (3a))

The majority of response types to kia questions are full predicates or sentences. This suggests that kia questions should be classified primarily as alternative or disjunctive questions, in which case the question asks whether or not X or Y is true. This categorisation concurs with our analysis of the pragmatically marked nature of kia questions. They are used in a variety of ways, for example, to seek out additional information, as in (22). In these examples, Q refers to the questioner and P refers to the Participant in the interview.

1. (22) Q: Ki Hakupu.

   gl.p Hakupu

   ‘To Hakupu.’

   P: E.

   right

   ‘Right.’

   Q: Ō hui nī kia?

   go foot emph q2

   ‘Just walked?’

   P: Ō hui nī he mogo pogipogi to liliu

   go foot emph loc.c time morning fut return

   mai he mogo afiafi ka hili e aoga,

   dir1 loc.c time evening when finish abs.c school

   pihia nā pogi.

   thus emph tomorrow

   ‘Go by foot in the morning time and come back in the evening time after school, same tomorrow.’(NOFIP5)

A kia question can also serve as a leading question, as in (23).

1. (23) Q: Fēfē haau a kitiaaga ki a lautolu

   what gen.2.sg lig view gl.p pers 3.pl

   ia he magaaho nei, kua kua pihia

   dem loc.c time dem perf perf thus

   kia haau a manatu ha kua kitekite

   q2 gen.2.sg lig thought because perf look

   atu a koe ke he tau fānau

   dir3 abs.p 2.sg gl.c loc.c pl children

   he magaaho nei, nākai makaukau lahi a

   gen.c time dem neg determined very abs.p

   lautolu ka.ha ke pehē fakaako e vagahau

   3.pl right sbjv thus learn abs.c language

   Niue po.ke fakaaoga e vagahau Niue.

   Niue or use abs.c language Niue

   ‘What is your view about them at this time – do you feel that way because you look at the children these days (and) they are not very determined to, say, learn the Niue language or use the Niue language?’

   P: Mooli lahi, mooli lahi.

   true very true very

   ‘Very true, very true.’ (NMMIP5)

A kia question can end with po ke ‘or’, as in (24). This differs from nākai questions which take the more restrictive optional yes–no question tag, e po ke nākai ‘yes or no’, as illustrated earlier in (16).

1. (24) E, ti ko e vagahau pālagi nī kia tūmau po.ke …

   yes so pred c language English emph q2 always or

   ‘Right, so speak always in English or …’ (NOFIP5)

Kia questions have one other characteristic. They often co-occur with an emphatic particle. In (25), the kia question co-occurs with the emphatic particle nī.Footnote ¹³

1. (25) E, kua vagahau pālagi tei a.ia ka ko e

   yes perf speak English almost then but pred c

   tapu Niue ne finatu a koe ki ai ka

   church Niue nfut attend abs.p 2.sg gl.p there if

   pehē ko e magaaho nī he tau Niue, ko

   thus pred c time emph loc.c pl Niue pred

   e heigoa e vagahau ne fakaaoga e lautolu ke

   c what lig language nfut use erg.p 3.pl sbjv

   fakatutala ka hili e tapu? Vagahau Niue nī kia?

   talk when finish abs.c church language Niue emph q2

   (NOFIP5)

   ‘Yes, then speak English then. But the Niue church that you attend…if it is only the time for the Niueans, what is the language they use to speak when the church is finished? Just speak in Niuean only?’

3.3 Ka questions

Ka has the most syntactic flexibility of all Q particles. It can appear in post-predicate position, as in (26), sentence-finally, as in (27), and in a range of other positions, such as after a non-predicate fragment, in (28), or even on its own, in (29). Its most common position is sentence-final.

1. (26) ko e hā ne malona ai e na e, kua leva

   pred c why nfut broken there prt prt prt perf long.time

   ka e malona?

   q3 abs.c broken

   ‘Why is this broken, has it been broken for long?’(NMMIP5)

1. (27) e, pihia he kitia e koe ka?

   yes thus loc.c see erg.p 2.sg q3

   ‘Yes, is that how you see it?’ (NOFIP5)

1. (28) matini hila ka?

   machine electric q3

   ‘Electric machine, eh?’ (NOFIP5)

1. (29) P: Ti mamate e tau kulene ia haaku ai kitia

   so die abs.cpl grandparents dem gen.1.sg neg see

   e au.

   erg.p 1.sg

   ‘… and died, those grandparents of mine, I did not see them.’

   Q: Ka?

   q3

   ‘Really?’ (NOFIP5)

NEG can co-occur with ka, as in (30).

1. (30) Ai ka mohe a Moka he mohega tote.

   neg q3 sleep abs.p Moka loc.c bed small

   ‘Didn't Moka sleep in the small bed?’ (Massam Reference Massam2003: ex. (3b))

The latter characteristic would group ka and kia questions together. However, ka questions cannot occur in wh-questions, as illustrated in (31).

1. (31) *Ko hai ka e tagata i kō?

   pred who q3 abs.c person gl.p dem

   ‘Who is that person there?’ (Massam Reference Massam2003: ex. (6b))

This is the only respect in which ka questions are more like nakai questions than kia questions. The presence of a wh-word is a signal of the information-seeking status of this structure. This co-occurrence restriction might suggest that ka is not a proto-typical yes–no question marker. This fits well with available information on response types, described below.

Ka is pragmatically marked. It is typically found in spoken rather than written discourse, and of the three question particles, questions with ka are the most likely to receive a minimal response. Ka questions, like kia questions, often follow up on information presented in the prior discourse. Ka appears in rhetorical questions, as in (32a–b).Footnote ¹⁴

1. (32)

   1. (a) ‘Ko e ō mai ka mua i tahi?’ he hūhū mai he akoako.

      pred c go dir2 q3 2.du gl.p sea as ask dir2 erg.c pastor

      ‘“Finished for the day?”, asks our pastor.’

   2. (b) ‘Ko e tau ika nī ka ia ha mua

      pred c pl fish emph q3 that gen 2.du

      ne hī?’ he vagahau he Matua Fifine mo e malimali.

      nfut catch as language gen parent woman com c smile

      ‘“Is this all you caught?”, asks Mum with a smile.’

It also appears as a follow up to a previous question or statement, as in (33).

1. (33) Q: Ka e iloa nakai e koe ka pehē ko e Tonga?

   but lig know q1 erg.p 2.sg if thus pred c Tongan

   ‘But would you know if it is a Tongan?’

   P: Kehe foki ha Tonga leo.

   different also gen.p Tongan voice

   ‘Tongans’ pronunciation is also different.’

   Q: E, iloa e leo Tonga ka?

   yes know abs.c voice Tongan q3

   ‘Yes, you know the Tongan pronunciation, eh?’ (NOFIP5)

The difference between ka and kia appears to be that whereas a ka question follows up on the previous question with no expectation that the situation is not true, a kia question appears to question the validity of information presented in previous turns. Therefore, although kia and ka form marked structures, the former is information-seeking whereas the latter is not. This would explain why the former, but not the latter, can appear in wh-questions. The final difference between kia and ka is that ka can appear on its own, where it appears to function as a discourse particle, as in (29) above and (34). The discourse particle status of ka is also evident from its preferred syntactic position in sentence-final position.

1. (34) P: Ko e Niue e matua taane ka e fanau i Ausetalia.

   pred c Niuean abs.c parent male but lig born loc.p Australia

   ‘The father is Niuean but born in Australia.’

   Q: Ka.

   q3

   ‘Eh?’(NOFIP 5)

Table 4

Summary of the distribution of the question particles. Shading marks differences between the particles.

3.4 Summary of yes–no question particle properties

The data thus far would suggest that nakai is associated with unmarked questions. It occurs in neutral contexts with a full range of response types. It can only occur in a core sentential, post-predicate position. In terms of typological classification, one could argue that Niuean does not have three yes–no Q particles. There is only one true unmarked yes–no question particle, nakai. Kia typically forms questions that have a negative bias, whereas ka serves as a discourse or confirmation particle, albeit one that usually occurs in a core sentential position, i.e. post-predicate, rather than in a typical peripheral position of a discourse particle. Table 4 outlines the distribution of the three forms presented thus far. From a syntactic point of view, it could be argued that nakai appears in a fixed position, whereas kia and ka can adjoin to other elements to provide various focus interpretations. We will now leave aside the issue of the syntactic and pragmatic status of the particles, and focus on their potential origins.

4.1 Kia questions

The particle kia occurs in yes–no questions as shown above, but it also appears in mild imperatives or exhortatives (Seiter Reference Seiter1980, Sperlich Reference Sperlich1997), as illustrated in (35).

1. (35) Kia hau a koe ki hinei!

   exhrt come abs.p 2.sg gl.p here

   ‘(You) come here!’(Sperlich Reference Sperlich1997: 153)

It is likely that both uses date back to the Proto-Polynesian form (PPn. *kia subjunctive). The original subjunctive usage would help account for both the strong attitudinal component expressed in the kia imperative and the pragmatic overtones evident in both kia questions and kia imperatives.Footnote ¹⁵ Of the Niuean structures, we assume that the mild imperative structure developed first. This proposed trajectory is supported by comparative evidence from Pukapukan, where the subjunctive ke < PPn. *kia is used to express commands but not yes–no questions (Salisbury Reference Salisbury2002: 539) and New Zealand Maori, where kia is an attitudinal TAM marker with imperative meaning (Bauer Reference Bauer1997: 99), but not a Q particle.

Another key component of our argument relates to the Principle of Persistence, where semantic features remain visible in subsequent structures. Seiter (Reference Seiter1980: 59) states that kia imperatives differ in pragmatic force from regular imperatives in that they ‘imply that the speaker has no control over the addressee’. As such, kia imperatives lack a key feature associated with other typical imperative structures. Kia questions are also pragmatically marked. Sperlich (Reference Sperlich1997) has noted that kia questions typically demand an answer. Thus, kia imperatives are less imperative-like than other imperatives, and kia questions are more imperative-like than either a nakai or a ka question.

In order to consider this claim, we need to examine the structures in more detail, all of which are present in modern-day Niuean. As stated earlier, in prototypical imperative structures, the subject noun phrase is deleted, and the clause has no TAM. When the subject NP is present, an imperative is interpreted as an offer or suggestion, as in (9) above. In other words, the pragmatic force of the imperative is reduced with an overt subject.

Kia imperatives differ structurally from offers and suggestions by the addition of the particle kia, as illustrated in Table 5, with a further shift in pragmatic force. Kia imperatives thus differ from kia questions in their syntactic position, clausal negation, and TAM. The latter two features, type of negative particle and the TAM marking, are dictated by clause structure. The only anomaly concerns the syntactic position of the particle kia.

Table 5

Differences between imperatives, requests, exhortatives and kia questions.

Several factors are likely to have played a role in the positioning of kia in yes–no questions. Central to the argument is the strong attitudinal component associated with the kia particle. Attitudinal markers are typically realised in Niuean as post-verbal particles (i.e. lā, nī, etc.), where they occur in the right periphery of the predicate (see Table 2 above). Attitudinal particles tend to co-occur, and it is not uncommon to have two or three such particles in a clause (Seiter Reference Seiter1980: 24). One common particle in imperative structures is the attitudinal particle lā. The Niuean language website (learnniue.co.nz) provides examples of lā in imperative structures, as in (36)–(37). Although Seiter (Reference Seiter1980: 58) does not provide any examples, he notes that kia imperatives typically co-occur with the emphatic particle lā.

1. (36) Ala lā.

   waken just

   ‘Wake up.’(learnniue.co.nz)

1. (37) Fano lā.

   go just

   ‘Go please.’(learnniue.co.nz)

Since both kia and lā have a strong attitudinal component and both regularly co-occur in mild imperatives, it is not difficult to see why kia may have been re-interpreted grammatically as an attitudinal particle. Once re-interpreted, kia would be free to co-occur with other attitudinal particles and this would provide the necessary conditions for kia to attract syntactic patterning similar to other attitudinal particles, and in particular the particle lā. Once this connection is made, other factors fall into place. There is a syntactic shift of kia to post-verbal position in affirmative clauses, and the adoption of an alternative pre-verbal order in negative clauses. The position of kia after lā would be expected if the former particle is a newer addition, as later particles are known to attach as peripheral elements. A later and weaker link might also help to account for why the Q particle kia can appear both after the verb and in sentence-final position.

Positing the origin of kia as an attitudinal marker accounts for a wide range of additional features in Niuean. It would explain the acceptability of kia in wh-questions, and its frequent co-occurrence with other emphatic markers. It would also account for the acceptability of multiple yes–no Q particles (Massam Reference Massam2003). The shift from an attitudinal marker to a question marker would be the next natural outcome, and would also have other ramifications, for example, once kia is interpreted as a question particle, its post-predicate position would be affirmed if a Q feature causes movement of the predicate into a left peripheral CP domain position, as suggested above. This movement would not take place in exhortatives (and optionally in negatives), leaving kia in pre-verbal position in such sentences. A summary of the development is presented in (38).

1. (38) Development of kia as a question particle

   • • Proto-Polynesian subjunctive kia developed a mild imperative (command) use with an attitudinal component (compare the Pukapukan subjunctive used for commands).

   • • The mild imperative frequently co-occurs with the post-predicate attitudinal particle (l ā).

   • • The co-occurrence of particles with similar semantics in similar structures creates the first set of conditions for a recategorialisation of kia as an attitudinal particle.

   • • A second structure which employs imperative verbs (but without the kia particle) is suggestions/requests. These structures are less imperative-like as they place more importance on the hearer, and are thus more question-like.

   • • Requests/suggestions differ from imperatives in that they have an overt subject NP, reaffirming VSO word order and the post-predicate positioning of attitudinal particles, creating the second set of necessary conditions for the recategorialisation of kia.

   • • Together, the two structures create the necessary ambiguity for the recategorialisation of kia as an attitudinal marker and the creation of a pragmatically marked question construction in Niuean (which demands an answer).

The reanalysis of kia as a Q particle with marked semantic uses would create a need for an unmarked question marker. If kia questions developed earlier than nakai, the development of an alternative word order for kia in affirmative and negative clauses could serve as a catalyst for the development of nakai as a post-predicate question marker. We turn now to discuss the origin of nakai as a question particle.

4.2 Nakai questions

The most likely origin of the nakai Q particle is the Niuean negative particle nākai. In this section we will argue for the development, outlined in (39), of this lexical item from being a NEG particle to being used also (with a slightly different form) as a Q particle.

1. (39) Development of nakai as an unmarked Q

   • • Proto-Oceanic NEG weakens and the resulting Polynesian languages develop alternative ways of strengthening NEG.

   • • The resulting Niuean form is the past tense marker (*na) and NEG (*kai). The past tense marker (na) also continues to occur as a TAM marker in Niuean, where it signals past uncertainty. The vowel lengthens (for unknown reasons).

   • • The weakening of NEG continues with structural consequences, with TAM continuing to attach to NEG but the post-predicate particles, with the exception of lā and tuai, now attaching to the predicate rather than NEG.

   • • The marked Q particle kia establishes a link between NEG and post-predicate position and creates a semantic niche for an unmarked Q particle.

   • • The semantic relationship between the irrealis markers NEG and Q allow the NEG particle nākai to fill this gap, and typological considerations re-inforce the preverbal position of NEG. The post-verbal position of Q is due to analogy with the marked Q particle kia and the typological need for Q and NEG to be distinct.

   • • When nakai occurs in post-predicate position as Q, it attaches as a clitic and loses a mora, further distinguishing NEG from Q.

We are positing that the Q particle nakai developed from the negative particle nākai. As the negative particle nākai is unknown outside of Niuean, we provide a little background on the possible origin of this particle. Lynch, Ross & Crowley (Reference Lynch John and Crowley2002) argue that the original negative forms in Proto-Oceanic lost some of their semantic force, and were subsequently reinforced. The reinforcements later replaced the original negator (in the manner of Jespersen's Cycle, Jespersen Reference Jespersen1917) or, in some cases, resulted in discontinuous forms (see e.g. Hooper Reference Hooper1993 on Tokelauan). This natural process helps to account for the diverse range of forms in Oceanic and by default the present-day Polynesian forms. In some of the daughter languages, NEG is derived on PPn *kai (Neg imperative) and phonological mergers between TAM and NEG (Clark Reference Clark1976, Salisbury Reference Salisbury2002). The TAM markers may have served as strengtheners as it is relatively common for languages to add a strengthener (see Dryer Reference Dryer, Hammond, Moravcsik and Wirth1988; Whaley Reference Whaley1997: 227–228) because negatives carry an important message, which needs to be unambiguously presented to the hearer. The Niuean NEG consists of the PPn TAM marker *na (past tense) and the PPn negative morpheme *kai ‘negative past tense’ (Clark Reference Clark1976).Footnote ¹⁶ A likely explanation for the Niuean form is that the past tense marker would have added greater uncertainty to the negative event or state. In Niuean, na refers to either a past event with some uncertainty or to an ongoing truth (Seiter Reference Seiter1980: 4), and this added marker of uncertainty may have contributed to a connection between NEG and Q. As this merger appears to be unknown outside of Niuean, we can assume that the NEG particle is a relatively recent historical change. If na were originally used to strengthen kai, this would entail that there had been a natural weakening of NEG. As the semantic force of NEG weakens, it is possible that its use could extend to other irrealis states and events.

Comparative data provide useful pragmatic information about Q which supplements the ideas presented above. Negative yes–no questions are used as requests and as invitations in some Polynesian languages (e.g. Pukapukan, see Salisbury Reference Salisbury2002: 499), and this could have been the case in pre-Niuean. If requests and invitations were originally formulated in Niuean with negative constructions, such a link would contribute to the potential use of NEG as a yes–no question marker in Niuean. The additional modern-day supplementary means of asking questions, by using the expression, e po ke nākai ‘yes or no’ illustrated in (16) above, would have further supported this connection.Footnote ¹⁷

Similarities between NEG and Q are well-documented typologically due to their shared semantic properties, and this may have also been a contributing factor. Both negatives and questions are non-veridical (Giannakidou Reference Giannakidou1998), being concerned with polarity, and dealing with uncertain or unrealised events or states, and they modify truth conditions of the embedded proposition (Horn Reference Horn1989, Horn & Kato Reference Horn, Kato, Horn and Kato2000a, among others). This semantic similarity has structural outcomes. For example, in English both yes–no questions and negation licence polarity items such as any. In other languages, the two often have similar forms: (Vietnamese khong, see Feuillet Reference Feuillet, Boucher and Fournier1994,, Inga chu see Whaley Reference Whaley1997: 227),Footnote ¹⁸ similar behaviour (do-support in English), or cannot co-occur, as they occupy the same underlying or surface syntactic slot (Québec French, see Morin Reference Morin2006, Reference Morin, Aboh, van der Linden, Quer and Sleeman2009). In Paduan, word order inversion appears in affirmative yes–no questions, but not in the negative, arguably because the two occupy the same slot, and are thus incompatible (Zanuttini Reference Zanuttini1997).

A similar, but not identical, argument could be put forth for Niuean. As with Québec French, the yes–no Q particle and the negative morpheme cannot co-occur, as shown in example (14) above. Although NEG and Q appear in different surface slots, the co-occurrence restrictions suggest that the forms for NEG and Q are linked in some fundamental way and we contend that this is because of their common origin historically, and possibly also their common merge position synchronically (Laka Reference Laka1994), assuming, as suggested above, a syntactic analysis in which the predicate moves to a position to the left of the Q marker, but does not move in case of negation. Negation and questions are not totally identical as there is no known language in which the two are not distinct, with negation tending to appear before the verb in VSO languages (Dryer Reference Dryer, Hammond, Moravcsik and Wirth1988, Horn Reference Horn1989, Feuillet Reference Feuillet, Boucher and Fournier1994). All of these examples are evidence of Structural Persistence of the vestiges of the original.

A recent development suggests the link between Q particles and focus exists in some languages but not others (see Ginsburg Reference Ginsburg2008). The latter may be of relevance here. Since sentential negation is associated with irrealis mood, Zanuttini (Reference Zanuttini1997) and Portner & Zanuttini (Reference Portner, Zanuttini, Horn and Kato2000), among others, have attempted to link such structures together, assigning Tense–Aspect–Mood and Negation to a NEG Phrase. Zanuttini has also presented an interesting argument, that incompatability constraints between negation and questions, and between questions and imperatives, show that the structures are closely aligned.There is an important syntactic difference between Q and NEG that needs to be explained. As illustrated in Tables 1 and 2 above, Q follows the predicate in Niuean, whereas NEG precedes the predicate. To explain this, we return to the assumption that NEG had verb-like properties in Proto-Polynesian and had TAM marking, post-predicate particles and complements, as is the case in a number of contemporary Polynesian languages (see Chung Reference Chung1970, Payne Reference Payne and Shopen1985). As the status of NEG weakens and it loses many of its verbal properties, including its ability to take TAM markers (perhaps because the TAM status is filled by the past uncertain particle na), and to attract a full range of post-predicate particles, there are consequences on the grammatical system. While the TAM markers continue to occupy a fixed position directly before the verb in negative clauses (TAM NEG V), the weakening of the verbal status of NEG results in an inability to attract post-predicate particles, with post-predicate particles now occurring in post-predicate position in both affirmative and negative clauses, with only the post-predicate particles on the far rightmost periphery remaining attached to NEG. As noted earlier in the paper (Section 2), this results in the rightmost particles, tuai and lā, occupying a post-verbal position in affirmative clauses and a pre-verbal (post-negative) position in negative clauses (see examples (4) and (5) above). The Q particle kia also now occurs to the right of the particles tuai and lā and occurs in negative and positive questions. The existence of alternative positions of the post-predicate particles tuai and lā in negative and affirmative structures and the ability for the marked Q particle kia to appear in its usual post-predicate position, as in (17) and (19) above, or to attach to NEG and occur in pre-verbal position, as in (21) above, creates the requisite surface conditions for NEG to occupy post-predicate position in its change from a NEG to Q.Footnote ¹⁹

Typological considerations further support this view of change. VSO languages have a tendency to place NEG in pre-verbal position (Dahl Reference Dahl1979, Dryer Reference Dryer, Hammond, Moravcsik and Wirth1988), which may well have helped nakai keep its syntactic position. Although in VSO languages Q also tends to occur in clause-initial position, there is a clear typological need for negative forms to be distinguished from question markers (Horn Reference Horn1989). Feuillet (Reference Feuillet, Boucher and Fournier1994) even goes so far as to claim that there is no known language in which NEG and Q are identical in form and structure. This need to avoid potential ambiguity is likely to have further supported the differing syntactic positioning and a further change. NEG and Q have a similar but not identical form (nakai vs. nākai). The difference between NEG and Q is likely due to the typological need for NEG and Q to be distinguishable. The difference in mora between the Q and the NEG is likely due to their different positioning in the clause. The Niuean Q particle attaches as the rightmost constituent of the verb phrase, which is a clitic-like position, whereas the negation marker does not cliticise, but appears before the predicate. Clitics are known to differ from free particles in a number of respects, including prosodic integration and a reduction in form (Zwicky Reference Zwicky1985, Aikhenvald Reference Aikhenvald, Dixon and Aikhenvald2003, Stebbins Reference Stebbins2003). A reduction in mora (vowel length) would be a natural outcome of grammaticalisation, which often co-occurs with phonetic reduction. The development proposed for nakai as a Q particle is outlined above in (39), at the beginning of this section.

4.3 Ka questions

In this section we argue for the development of the Q particle ka from a discourse particle, as outlined in (40).

1. (40) Origin of ka as a question marker

   • • Kaha is a fixed verbal phrase, which appears in a wide variety of interrogative and non-interrogrative contexts.

   • • Kaha is reinterpreted as a discourse particle due to its frequent sentence-final position.

   • • As a discourse particle, kaha attaches to S (sentence), where it was prone to phonetic reduction; kaha reduces to kaː

   • • As a grammatical particle, kaː is subject to a phonological rule: Word-final monosyllabic words with grammatical functions have short vowels (kaː – ka).

   • • Homophonous forms with irrealis functions (e.g. ka ‘future, but, if’) reinforce the irrealis status of ka.

   • • In discourse, post-predicate position often equates with sentence-final position.

   • • Once these links are drawn, further links are made between the homophonous morpho-phonemic variant of kia [ka] and its syntactic positioning in affirmative and negative clauses.

Both observable phonetic synchronic varability and the absence of ka as a Q particle in contemporary Niuean grammars (Seiter Reference Seiter1980) suggest that this question particle is a very recent development. The only mention of any Q particle ka in Seiter's grammar is as a morpho-phonemic variant of the Q particle kia. Although this morpho-phonemic variant is likely to have played a role in the development of the ka question particle, discussed below, it is not the likely origin of the particle.

The most likely origin of ka is the lexical item kaha ‘that's it/that's right’. The two items have similar semantics, providing or requesting confirmation from the listener. The confirmational phrase origin also helps explain syntactic restrictions on ka questions. As stated earlier (Section 3.3), ka cannot occur in wh-questions, which are information-seeking rather than confirmational. Another piece of evidence for this origin comes from the answers to ka questions, which are largely confirmational responses rather than yes or no responses.

As a lexical item, kaha ‘that's right’ is frequent in the data, and, as illustrated in (41), it sometimes has no status as a question particle. It serves simply as a sentential tag.

1. (41) Maloloː na au ne nofo ai fano he aoga

   well pst 1.sg nfut stay neg go loc.c school

   a ia mo e tau gahua lima ha matulu

   abs.p 3.sg and lig pl work hand gen.c1.excl.pl

   tau community kaha

   pl community right

   ‘Well I was the one who stayed home and did not go to school, we worked our handicrafts, that's right.’ (Massam et al. Reference Massam, Starks, Ikiua, Moyse-Faurie and Sabel2011: ex. (26))

At other times, as in the example in (42), kaha appears in structures in yes–no questions.

1. (42) Q: ne mukamuka a koe kaha?

   pst easy abs.p 3.sg right

   ‘because it is easy for you, is that right?’ (Massam et al. Reference Massam, Starks, Ikiua, Moyse-Faurie and Sabel2011: ex. (25))

   P: Yeh

   ‘Yeah.’

It is likely that an alternative means of forming yes–no questions, by intonation, is acting as the marker of Q here. Examples such as (43), where kaha co-occurs with ka, strongly suggest that kaha, when it appears in Q structures, does not signal a yes–no question.

1. (43) Uafulumahiva e tau a ia haau kaha ka?

   twenty-nine lig year abs.p 3.sg gen.2.sg right Q

   ‘You were twenty-nine years then, right?’

There is ample evidence that the phonological reduction in the form of the yes–no question particle is a recent change. The particle appears in a range of phonetic forms in the speech of the Niuean interviewer in the Pasifika Languages of Manukau Project. The variants form a synchronic cline with the bisyllabic kaha at one end and the monosyllabic Q particle ka on the other. Between the two ends are two additional forms, kaː and kā.Footnote ²⁰ The form kaː contains a long vowel that is further back, probably due to the loss of the segment [h]. The reduction in mora with compensatory backing would be a natural grammatical process when a lexical item gradually loses its independent status and becomes a clitic particle. The other variant is kā, with a long mid-front vowel. Both forms are relatively infrequent. The most common variant, ka, is a natural outcome of grammatical constraints on vowel length in Niuean, where all monosyllabic grammatical items show no contrast in vowel length. The short vowel provides strong evidence that the question particle is now fully grammaticalised. The co-occurrence of kaha and ka, as in example (43), would also support this claim.

Once kaha came to be phonetically reduced, the question particle came to resemble other particles in Niuean that have an irrealis function: ka ‘but’, ka ‘if/when’ and ka ‘future subordinate’. One of these irrealis particles occurs in the fixed expression ka i ko NP ‘but what about NP’, as illustrated in (44), and is relatively frequent in conversation.

1. (44) Ka i ko e haau a tama?

   but cause pred c gen.2.sg lig child

   ‘What about your child?’(NOFIP5)

The structure is used to introduce intonation-based questions, which follow on from other discourse in ways somewhat similar to those of ka questions. A second construction, with a clause-initial irrealis marker, ka + S ‘but S’, also occurs frequently in conversation, where it functions as a discourse marker. In this context, it can introduce yes–no questions, as in (45).

1. (45) Ka e iloa nakai e koe ka pehē ko e Tonga?

   but lig know q1 erg.p 2.sg if thus pred c Tongan?

   ‘But would you know if it is a Tongan?’(NOFIP5)

Both of these structures might have helped to create a link between ka as a confirmation marker and ka as Q particle.

The yes–no Q particle, ka, does not appear in clause-initial position. It occurs in post-predicate position in affirmative clauses and pre-predicate position in negative clauses. It also attaches to NPs, occurs on its own, and appears in sentence-final position. The latter syntactic positioning can be explained, in part, as derived from kaha. As a single stand-alone lexical item, kaha appears on its own, and as a confirmational tag, it often occurs in clause-final position. Its other positions, as a post-predicate particle in affirmative clauses and as a pre-predicate particle in NEG clauses need further explanation. As mentioned above, post-predicate position often equates with sentence-final position, where the following arguments are not overtly realised in the discourse context, providing ambiguity in learner input. The frequent use of a sentence-final post-predicate Q particle would provide ambiguous models for language learners, potentially leading to its use in post-predicate non–sentence-final constructions.

The syntactic positioning is likely triggered by another factor. Both ka and kia questions share similar phonological form under certain morpho-phonological conditions. Seiter (Reference Seiter1980) notes that when kia occurs in post-predicate position and is followed by the case marker a, it is realised as ka. This phonetic similarity between this morpho-phonemic variant of kia and the particle ka could have created the necessary ambiguity to promote the use of ka in post-predicate non–sentence-final position.

This extension of ka from sentence-final to non–sentence-final post-predicate position would result in all three Q particles in the Niuean language occuring in post-predicate position. Once this association is made, it would be fairly certain that the pre-predicate positon of ka in negative questions would be a natural consequence. The trajectory is sketched above in (40), at the beginning of this section.