1 Introduction

Every language has a way to negate a declarative sentence (Dahl 1979, Payne 1985, and Dryer 2005). For sentential negation, three main types are attested: morphological negatives, negative particles, and negative verbs (auxiliary or matrix).¹ Negative particles are understood as being within a negative phrase, or NegP, and can be heads or adverbs (Pollock 1989 and Zanuttini 1997). NegP appears in the clausal spine and selects either a TP or VP complement (Ouhalla 1990). In contrast, negative verbs are themselves predicates, have their own tense/aspect markers, and take a sentential complement (Zanuttini 2001).

It has been argued that Polynesian languages have predicate negation, with a negative predicate followed by a subordinate declarative clause. Most famously detailed by Chung 1970, this analysis has been adopted for Māori (Hohepa 1969, Chung 1978, Waite 1987, and Chung 2021), Niuean (Clemens 2018), Tongan (Broschart 1999, Custis 2004, and Ball 2008), Tahitian (Lazard & Peltzer 1999, Potsdam & Polinsky 2012, and Potsdam & Polinsky 2017), East Futunan (Moyse-Faurie 1999), Tokelauan (Hooper 1993 and Vonen 1999), and Samoan (Mosel & Hovdhaugen 1992). Examples of negation from these languages are the following.²

However, there are also claims that Polynesian languages exhibit particle negation instead. Some authors suggest that some Polynesian languages contain both predicate and non-predicate negation (e.g., Hooper 1993, Hovdhaugen & Mosel 1999, Vonen 1999, Custis 2004, and Ball 2008) while others argue for exclusively particle negation (Collins 2017 and Middleton 2023). For some languages, there are conflicting claims on the category of certain negative words, such as Niuean nākai (Clemens 2018 vs. Massam 2020), Tongan ‘ikai (Broschart 1999 vs. Custis 2004), and Samoan lē and le‘i (Chung 1970 vs. Collins 2017).³ It is clear from this that negative structures in Polynesian languages continue to be debated. Consequently, diagnostics that can differentiate between particle and predicate negation are vital for a fuller understanding of negation strategies in these languages and for a comprehensive description of the negative predicate category proposed by Payne 1985.

There are several diagnostics that have been employed to determine whether a language has predicate negation, including secondary tense/aspect/modal (TAM) particles, a subordinating complementizer, the presence of predicate modifiers next to the negative, and raising of an argument from the verb to the negative (Hohepa 1969, Chung 1970, Chung 1978, Bauer 1997, Pearce 1997, Custis 2004, Ball 2008, Potsdam & Polinsky 2012, Potsdam & Polinsky 2017, Clemens 2018, and Chung 2021). These are briefly illustrated below.

In Māori, as shown in (8), negatives are followed by a second TAM particle, in addition to the preceding one. Note that in this article, square brackets always denote clause boundaries.

It is argued that the first TAM particle is for the negative predicate while the second is for the subordinate verbal clause. In fact, only some TAM particles are ever used to introduce subordinate clauses, and it is only these TAM particles that are ever seen following negatives (Chung 2021).

Instead of a secondary TAM particle, Tongan has a complementizer particle, ke, that follows certain negative particles, as in (9b). Like in Māori, this indicates that the negative is followed by a subordinate clause.

Māori negatives may also be followed by verbal particles, indicating that they are verbal predicates. For example, a negative can be followed by manner and directional particles, such as noa ‘freely’ and ake ‘away,’ as in (10a). These are not simply modifying the verb to the right, since the verb itself can be concurrently modified by such particles, as in (10b).

In Tahitian, negative predicates are raising verbs. The subject of a verb must obligatorily raise to the negative predicate, raising above the verb’s TAM particle and into the higher clause. This can be observed with the second person singular pronoun in (11b).

An additional example, from Māori, is seen in (10b), with the subject he tangata o te iwi nei ‘a man of the tribe’ raising from the verb noho ‘sit.’

Together, these pieces of evidence support the analysis of negatives in these languages as predicates. However, these diagnostics are unidirectional; the presence of these particles/processes provides evidence for negative predicates, but the absence of them does not provide evidence against predicate negation. In spite of this, the absence of these particles/processes has often been taken as evidence for particle negation in Polynesian (e.g., Massam 2000 and Custis 2004).

It is clear that we do not have adequate diagnostics for negation in Polynesian languages and that there is a need for diagnostics that argue for or against predicate negation. This article proposes that apparent raising (Otsuka 2000, Polinsky 2016, Otsuka 2018, and Otsuka 2021) can be employed to identify predicate negation or rule it out. To illustrate this diagnostic, I turn to Tokelauan and Samoan (both Polynesian, Samoic).⁵ Raising verbs in these two languages allow an argument from a subordinate clause to raise up to the matrix clause. Only certain verbs allow this movement; non-raising verbs do not. An example of apparent raising, from Tokelauan, is the following.

A string of raising verbs allows long distance raising of an argument from the lowest clause to the highest:

However, if a non-raising verb intervenes in this string of raising verbs, the argument is blocked from raising past it. In (14), although vaganā ‘unless’ is a raising verb, the non-raising verb lea ‘say’ blocks raising from that clause.

I argue that if negation is a predicate, it will either be a raising predicate and therefore allow raising to it or will be a non-raising predicate and therefore block raising past it. If negation neither allows raising nor blocks it, as I illustrate is the case in Tokelauan and Samoan, it cannot logically be a predicate and therefore must be a particle instead.

This article is ordered as follows. Section 2 introduces Tokelauan and Samoan morphosyntax. Section 3 discusses negation in Polynesian and in particular the absence of diagnostics for particle negation. Apparent raising is discussed in section 4, where the pattern of long distance raising is illustrated in more detail along with the blocking of long distance raising by an intervening non-raising predicate. Section 5 demonstrates that apparent raising can be employed as a diagnostic for particle negation: this diagnostic shows that Tokelauan and Samoan negatives can neither be raising predicates nor non-raising predicates. Section 6 concludes.

2 Tokelauan and Samoan morphosyntax

This article will examine Tokelauan and Samoan, both of which fall in the Samoic subgroup of Polynesian (Hovdhaugen et al. 1989). Tokelauan has 3,900 speakers, while Samoan is spoken by 400,000 people (Simons & Fennig 2017).

Both languages are predicate-initial and have an unmarked VSO word order, with a VOS order also possible for both (Mosel & Hovdhaugen 1992 and Middleton 2024).⁶ They have ergative–absolutive case alignment and the same case markers; e is obligatory for ergative arguments while absolutive arguments are unmarked, except R-expressions, which are preceded by the absolutive marker ia. These facts are illustrated in (15) and (16).

Verbs are usually preceded by a TAM particle, as (17) illustrates, although complementizers can replace the TAM in subordinate clauses, as in (18). Imperatives, as (19) illustrates, have no TAM or complementizer. Pre-verbal pronoun clitics are sometimes seen preceding verbs, as in (17b).

Negatives in Tokelauan differ depending on aspectual distinctions; hēki is used for events and hē for states. Negatives follow the TAM particle and precede the main predicate constituent, as in (20). Negation triggers the -cia suffix on transitive verbs, as (20a) illustrates.

As (21) shows, Samoan also has two negative particles, which precede the verb: lē (which has an allomorph le) ‘not’ and le‘i ‘not yet’ (Mosel & Hovdhaugen 1992: 142).

Both languages exhibit apparent raising (Otsuka 2000, Polinsky 2016, Otsuka 2018, and Otsuka 2021), a process where an argument from a subordinate clause may optionally raise to the matrix clause, sometimes (though not always) exhibiting a different morphological case marking:

Apparent raising will be discussed in detail in section 4.

3 Negation in Polynesian

There is much evidence to support a predicate analysis of negation in Polynesian languages. Among the diagnostics are the presence of two TAM particles in the sentence, the presence of a subordinating complementizer following the negative, verbal modifiers on the negative, and the ability of an argument to raise from the verbal clause to the negative clause (Hohepa 1969, Chung 1970, Chung 1978, Chung 2021, Bauer 1997, Pearce 1997, Custis 2004, Ball 2008, Potsdam & Polinsky 2012, Potsdam & Polinsky 2017, and Clemens 2018). Examples were given in the introduction (8–11).

However, there is an absence of empirical support for a non-predicative analysis of Polynesian negation, despite the many authors who propose just that (Hooper 1993, Hovdhaugen & Mosel 1999, Vonen 1999, Massam 2000, Custis 2004, Ball 2008, Collins 2017, and Massam 2020). The evidence presented for particle negation has usually been a lack of evidence for predicate negation: if a negative word does not exhibit clear signs of being predicative, it is assumed that this negative word is a particle in NegP.

For example, Tongan has both negation followed by a complementizer ke, as in (24a), and negation followed directly by the verb with no complementizer, as in (24b).

Custis 2004 proposes that these are different forms of negation: that the former is a predicate while the latter is a non-predicative particle. This is based predominantly on the lack of the complementizer in (24b). Ball 2008 takes an opposing view and claims that both forms are predicates but that the first selects a TP complement while the second takes a vP complement.¹¹ I suggest that both proposals are fundamentally the same: (24a) contains a matrix negative predicate with a subordinate verbal clause, while (24b) has something more akin to a negative auxiliary or NegP, which slots into the clause below TP and above vP. Crucially for this article, the diagnostic for the latter is simply the absence of lexical material, rather than anything more substantial.

Massam 2000 proposes that the Niuean negative nākai is a particle (cf. Clemens 2018), but the reasons are equally unsatisfactory. Massam claims that the complement of nākai takes what appears to be a VP complement, not a CP complement, as it is not followed by either a complementizer or TAM particle. In addition, nākai cannot host a full range of post-verbal particles as would be expected of a verb. Both of these diagnostics are simply the absence of certain things expected with predicate negation.

In order to gain a better understanding of negation in Polynesian, clear diagnostics are needed that can determine the syntactic category of negation. This article employs apparent raising as a diagnostic for both predicate negation and non-predicate negation.

4 Apparent raising

Apparent raising is a process where an argument in a subordinate clause is optionally raised to the matrix clause (Chung 1978, Seiter 1978b, Chung & Seiter 1980, Besnier 1988, Otsuka 2000, Polinsky 2016, Otsuka 2018, and Otsuka 2021). The verb of the higher clause is called a raising verb. Raising verbs are often (but not always) intransitive, and only certain verbs can trigger raising. Apparent raising differs from true raising (in the sense of Postal 1974) in that non-subjects can undergo raising and raised arguments can “skip” a clause and raise to the highest clause of a multi-clause sentence (Polinsky 2016).¹²

In both Tokelauan and Samoan, apparent raising appears to be Ā-movement, allowing unbounded movement across multiple clauses, as long as each higher clause contains a raising verb. In (25), the Tokelauan verbs vaganā ‘unless’ and mafai ‘to be able’ are raising predicates.¹³ Consequently, the subject of the verb mātamata ‘watch,’ namely ki tātou ‘we (incl),’ is able to raise from its position in the lowest clause, seen in (25a), to the second clause, as in (25b), and then again to the third clause, as in (25c). The interpretation of all three sentences is identical. In each case, ki tātou ‘we (incl)’ moves to a position above the complementizer or TAM particle.

In a similar fashion, mānatua ‘remember’ and masani ‘to be usual’ are raising verbs in Samoan, which allows the subject of the verb tagi ‘cry’ in (26), namely le pepe ‘the baby,’ to raise from the lowest clause, where it is found in (26a), to the second clause, as in (26b), as well as to the highest clause, as in (26c).¹⁴ Again, interpretations are identical.

Note, however, that successive raising can only occur if all the higher verbs are raising verbs, such as in (25), where both vaganā ‘unless’ and mafai ‘to be able’ are raising verbs, and in (26), where mānatua ‘remember’ and masani ‘to be usual’ are raising verbs. In both Tokelauan and Samoan, a non-raising predicate will block raising, even if a higher raising verb dominates the non-raising verb. In Tokelauan, mahani ‘to be usual’ is a raising predicate, but lea ‘say, tell’ is not. In (27), the presence of the non-raising predicate lea ‘say, tell’ blocks the subject of the lowest clause (te vaka ‘the boat’) from raising to the highest clause, even though the highest clause has a raising predicate.

Similarly, in Samoan, mafai ‘to be able’ is a raising predicate, but in (28) the non-raising predicate ta‘u ‘tell’ blocks the subject of the lowest clause (le pepe ‘the baby’) from raising to the highest clause.

Raised arguments may have different case marking compared with the non-raised variant. Hooper 1993 observes that ergative, absolutive, and occasionally prepositional nominals can be raised in Tokelauan, with the raised argument taking ergative, absolutive, or oblique morphological case, not necessarily the same case as in the non-raised version. Similarly, Samoan exhibits S-to-A, A-to-A, S-to-O, and O-to-O raising, where S is an intransitive subject, A is a transitive subject, and O is a direct object (Mosel & Hovdhaugen 1992).¹⁵ An example of ergative-to-absolutive (A-to-O) raising in Tokelauan is given in (29), and an example of S-to-A raising in Samoan is given in (30).

Apparent raising has been analyzed in two ways. In one model, advocated by Otsuka 2000 and Polinsky 2016, the raised argument is base-generated in the matrix clause and co-indexed with a null operator in the subordinate clause; this can explain the Ā-movement properties, such as long distance dependency (which true raising does not exhibit). The other analysis, advocated by Otsuka 2018 and Otsuka 2021, has the argument raised to the embedded specCP above the complementizer and assigned case from the matrix verb, with no null operator involved; when an argument raises to specCP, Otsuka proposes that the noun’s previously assigned case value is erased and the matrix verb instead assigns case to the raised noun. I remain agnostic about an analysis of apparent raising.¹⁶ In this article I use raising as a descriptive diagnostic, one that is therefore theory-neutral. Consequently, in what follows I will simply use the term raising to mean that there is apparent movement of a noun from a subordinate clause to the matrix clause, without committing to an analysis of how this occurs.

5 Apparent raising as a diagnostic for negation

Negation in many Polynesian languages has been argued to be predicative. If negation is a predicate, there are two options: negation is a raising predicate, or negation is not a raising predicate. The diagnostic discussed in this section hinges on proving that negation in Tokelauan and Samoan does not exhibit characteristics of raising predicates nor characteristics of non-raising predicates; it is thus proposed that negation is non-predicative and instead must be a particle found in the clausal spine.

The ability of an argument to raise up to the negative verb has been used as a diagnostic for predicate negation (Hohepa 1969, Chung 1970, Chung 1978, Bauer 1997, and Chung 2021). For example, in Cook Islands Māori, the subject of a negated stative verb can appear following the verb, as in (31a), or raise to the negative predicate, as in (31b) (see also (11) for a Tahitian example, as well as (10b) for a Māori example).¹⁷

I propose that raising can also be used to rule out predicate negation. If negation is a predicate, it must either be a raising predicate or a non-raising predicate. If negation in Tokelauan and Samoan is a raising predicate, we would expect an argument to be able to raise from a position following the verb to a position directly following the negative. This is ungrammatical:

Consequently, we can categorically state that negation in these languages is not a raising predicate, if it is indeed predicative. Recall that in this article, square brackets denote clause boundaries; in (32, 33), the square brackets reflect the assumption that negatives are predicates.

Having established that if negation is a predicate, it cannot be a raising predicate, we can now check whether negation is a non-raising predicate instead. Recall that in a three clause sentence, an argument from the lowest clause may be raised to the highest clause only if both higher verbs are raising verbs. A non-raising predicate, which we now assume negation to be, is expected to block movement of an argument from the lowest clause to a matrix clause with a raising predicate—but, as (34) demonstrates, this does not happen. In the Tokelauan example (34a), the subject of the lowest clause (te fafine ‘the lady’) is able to raise above the lowest verb fano ‘to go’ and the negative hē into the highest clause, which contains the raising predicate mahani ‘to be usual.’ Equivalently, in the Samoan example (34b), the subject of the lowest verb tagi ‘to cry’ (le pepe ‘the baby’) is able to raise above the verb and the negative le to the clause containing the raising predicate mafai ‘to be able.’¹⁸

This is evidence that if negation is a predicate, it does not act like a non-raising predicate, which would block raising past it. We now come to an impasse: negation does not behave like a raising predicate, nor does it behave like a non-raising predicate. Consequently, with no alternatives, negation cannot be a predicate. We may therefore conclude that Tokelauan and Samoan negation is a particle and that the NegP proposal, as in Custis 2004, Collins 2017, and Massam 2020, is more suitable for these two languages.¹⁹

6 Conclusion

This article provides a diagnostic that can be employed to identify predicate negation versus non-predicate negation (i.e., particle negation). While particle negation has been proposed for many Polynesian languages (Hooper 1993, Hovdhaugen & Mosel 1999, Vonen 1999, Massam 2000, Custis 2004, Ball 2008, Collins 2017, and Massam 2020), evidence for this analysis has been thin. The main argument provided is the absence of particles usually seen in predicate negation: for example, the lack of a TAM particle or complementizer following the negative word might be taken to indicate non-predicate negation. Unlike those diagnostics, this article proposes a diagnostic that either supports a negative predicate analysis or systematically rules it out. With differences in how negation has been analyzed in many Polynesian languages, diagnostics like these are vital for syntactic and typological research into negation in this family of languages.

While the diagnostic in this article is only illustrated for two languages, other Polynesian languages exhibit many of the same properties, which should allow the diagnostic to be used further afield. For example, apparent raising is also attested in Māori (Bauer 2005), Cook Islands Māori (Nicholas 2016), Tuvaluan (Besnier 1988), Pukapukan (Salisbury 2002), Niuean (Chung & Seiter 1980), Tongan (Otsuka 2000), and Rapa Nui (Kieviet 2017). The category of negation is debated in several of these languages, including Niuean (Clemens 2018 and Massam 2020) and Tongan (Broschart 1999, Custis 2004, and Ball 2008), and is unanalyzed in several of the others. I therefore hope that the present diagnostic has use across the language family and is able to better illuminate the complexities of predicate and non-predicate negation.

In addition to presenting this diagnostic, I have also proposed that Tokelauan and Samoan do not have predicate negation. This itself is noteworthy, since Samoan was one of the earliest languages to be analyzed as having predicate negation (Chung 1970) and the results from this article indicate the opposite. Further development in this topic will clarify the extent of predicate negation, which may be restricted to fewer languages than first imagined.

The diagnostic set out in this article makes a typological prediction of three options. Firstly, negative predicates in some languages may license apparent raising. In these languages, the raising negative predicate allows an argument in a subordinate clause to raise to the negative predicate clause, as shown in (35). This occurs in Tahitian (see (11)), Cook Islands Māori (see (31)), Māori (see (10b)), and Niuean (Seiter 1980).

Secondly, negation may be a particle, which will not interact with raising at all (as this article argues is the case for Tokelauan and Samoan). In these languages, an argument is able to raise above a negative to a clause containing a raising verb:

A third theoretical possibility is that negation is a predicate but does not license raising. In such a language, an argument from a deeply embedded clause should not be able to raise to the negative predicate clause nor to a superordinate clause containing a raising verb:

While we have no examples of a language of this third type, the diagnostic from this article predicts that such a language should be able to occur in the set of natural languages. I leave this for further research.

Acknowledgments

I thank Angela Sagala-Mila for sharing her language (Samoan) with me, Iutana Pue for his help with the Tokelauan judgments, and the two anonymous reviewers for their comments. All mistakes are my own.

Competing interests

The author has no competing interests to declare.