Revisiting the clause periphery in Polynesian languages

Verb-initial languages often contain a pre-verbal particle, which, in Polynesian languages, is a tense/aspect/modal (TAM) marker. For Tongan and Samoan, it is standardly assumed that TAM markers are generated in T˚, which are then moved to C˚ in current frameworks (T-to-C movement), meaning TAM and complementisers are in complementary distribution (Custis 2004; Otsuka 2005; Collins 2017). This squib presents novel data from Tokelauan, another verb-initial Polynesian language, showing that TAM particles and complementisers can co-occur, indicating that T-to-C movement is more complex than originally imagined. I propose that an expanded left periphery is needed, with two complementiser positions, and TAM raising to the lower one of these.


Introduction
The Linear Correspondence Axiom, which states that languages are underlyingly SVO (Kayne 1994), means analyses of verb-initial languages typically include raising the verb head (or phrase) to a position high in the clause, in order that the verb precedes the arguments in the surface form (Carnie & Guilfoyle 2000;Carnie et al. 2005). As noted by Bury (2005), verbinitial languages typically contain a pre-verbal particle, which must be accounted for when deriving the verb-initial structure. This squib makes the claim that in three Polynesian verbinitial languages, an expanded left periphery is required to account for the placement of this pre-verbal particle.
Many Polynesian languages exhibit a verb-initial syntax, along with a tense/aspect/modal (TAM) particle preceding the verb. Following the standard assumption that the TAM particle is base-generated in T˚ (Massam 2000;Otsuka 2005;Collins 2017), analyses of Tongan and Samoan require TAM to raise to C in order to derive the correct word order in those languages. In each language, it is claimed that an argument DP can (or must) raise to SpecTP. Having concluded that one or the other argument ends up in SpecTP, these analyses are forced to consider that TAM is raised, from T to C, in order to derive a TAM-initial word order. If TAM remains in T, the raised argument will not precede TAM, which is empirically incorrect. T-to-C movement solves this issue, and as such is proposed by Otsuka (2005) and Collins (2017) for Tongan and Samoan respectively.
Delving further into the C domain causes problems for T-to-C movement. If the CP is a single projection, co-occurrences of TAM and complementisers are not predicted by previous analyses of these languages. In this squib I will present novel data from another verb-initial Polynesian language, Tokelauan, and argue that a single CP projection cannot be maintained in order to derive the TAM-initial order in this language. 1 Critically, some complementisers and TAM markers can co-occur in Tokelauan, and do not constitute a singular morphological unit, which disallows the possibility of T-to-C movement through cyclic adjunction. Furthermore, I will provide parallel data from Tongan and Samoan that show that even in these languages, there are some complementisers that can co-occur with TAM, a fact that has been overlooked in the earlier works, which suggests that T-to-C movement in these languages should be revisited in the light of this data. I show the problem is the landing site of TAM, not the fact that it undergoes head-movement.
To rescue TAM-movement, I will propose an expanded CP (Rizzi 1997), with two positions for two different types of complementisers. I will suggest that TAM raises to the lower of these positions, resulting in the complementary distribution of these complementisers and TAM. The higher position is where the complementisers that can co-occur with TAM are generated. The suggestion of two different complementiser positions is supported by the distribution of fronted topics. In Tokelauan, as in many Polynesian languages, topics are ko-marked and raised to the left periphery. Ko-marked topics in Tokelauan are shown to follow the complementisers that I argue to be in a higher position, but precede the lower complementisers -suggesting that the two types of complementisers have different syntactic positions, one higher than the ko-marked topics, and the other lower.
This squib is structured as follows: Section 2 will review the literature on Tongan and Samoan, outlining the reasons to posit T-to-C movement in these languages. Section 3 will provide Tokelauan data, as well as additional parallel data from Samoan and Tongan, to show the T-to-C model needs revisiting. Section 4 will propose an account based on expanded left periphery to account for the data; further support for this analysis is provided from ko-topicalization in Tokelauan. Section 5 concludes the discussion.

T-to-C Movement
This section will briefly examine the word-order analyses of two Polynesian languages: Tongan and Samoan. Specifically, it will focus on why current analyses of these languages require T-to-C movement. 1 The novel Tokelauan data in this squib comes from work with two native Tokelauan language consultants, although many equivalent examples were found in Hooper (1993). Any data that is not referenced to another author is from this author's consultants.  Otsuka (2005) proposes an account for Tongan that includes T-to-C movement. Her analysis for the verb-initial order in the language is based in a head-movement approach, where it is proposed that the verbal head undergoes V-to-T-to-C movement. The reason for positing V-to-T movement is that Tongan exhibits argument scrambling, allowing both VSO and VOS sentences, as in (1a-1b) respectively.
Na'e fili 'a Pila 'e Sione. pst choose abs Pila erg Sione 'Sione chose Pila.' Following Chomsky (1993), Otsuka asserts that if V moves to T, the subject and object are equidistant from T˚. She proposes T˚ contains an EPP feature, which draws the closest DP to the specifier position of TP, under the Minimal Link Condition (Chomsky 1995). Since V-to-T movement renders both arguments equidistant from T˚, this allows either argument to raise to SpecTP. VSO clauses involve the subject DP raising to SpecTP, whereas VOS clauses involve the object DP moving. In either case, one DP is moved to the specifier of TP. Inorder to derive the verb-initial word order, the verb must raise higher that the DP in SpecTP, hence why Otsuka argues for the verb to move further to C.
While there are independent reasons to question the validity of the V-to-T-to-C movement proposed by Otsuka, such as the issue of leftward head adjunction, 2 the main point is that T-to-C movement is required in Otsuka's analysis to obtain TAM-initial ordering.
Another argument for T-to-C movement is that complementizers and TAM particles cannot co-occur together. Custis (2004) argues that this complementary distribution results from the TAM and complementisers competing for the same syntactic slot (C˚). This distribution is shown for Tongan in (2). 3 The complementiser 'o introduces the embedded clause, which does not, and cannot, contain a TAM particle. As all other verbal clauses contain a TAM particle, it is argued that this slot is already filled by the complementiser 'o, and thus no overt TAM particle is possible.4 (2) Tongan (Custis 2004: 120) Na'e ha'u 'a Mele 'o nej kaiha'asi 'a e siaine. pst come abs Mele comp 3sg steal abs det banana 'Mele came and stole the banana.' 4 Note the presence of "the pronominal subject" occurring pre-verbally, a common occurrence in Polynesian languages, which will play a significant role in the rest of this squib. Collins (2017) investigates Samoan verb-initial order and posits T-to-C movement for several reasons. First, Collins demonstrates that Samoan TAM markers appear to show properties 2

Samoan
Under the Mirror Principle (Baker 1985), V-to-T-to-C movement should form a complex head that has the linear order [V+T+C]. Tongan is a head-initial language, and so any head that undergoes cyclic head movement will attach to left side of the higher head. V-to-T-to-C movement is understood as cyclic adjunction, where V˚ adjoins to T˚ to form a morphologically complex head [V+T]. This in turn is moved to join onto the left of the head of C˚, forming [V+T+C]. This is not the order observed in Tongan sentences, which is instead Note that this squib will show later on that this complementary distribution between complementisers and TAM particles holds only with certain complementisers, but not all.

4
The original translation in Custis (2004) appears to be erroneous. This has been corrected in this squib. related to both T and C. 5 For instance, in relative clauses, a TAM particle precedes the relative clause, and therefore is interpreted as a relativizing complementiser.
(3) Samoan ( Furthermore, Collins notes that the complementiser ona is in complementary distribution with TAM particles. When a verb takes the complementiser ona, the embedded verb may not be preceded by any TAM marker. This leads Collins to argue that T˚ and C˚ are syntactically merged, having undergone T-to-C movement. 6 (4) Samoan (Collins 2017: 30) 'Ua siliga ona (*sā/e/'ā) taunu'u mai le tama. prf too.late comp (pst/prs/fut) arrive dir det man 'The man was overdue coming back.' (lit. it was too late that the man came back) In addition, T-to-C movement in Samoan is needed to understand the distribution of preverbal subject pronouns. Like Tongan, Samoan also has subject pronouns occurring preverbally in transitive sentences. Collins analyses these subject pronouns as moving to SpecTP from their base-generated position in vP. In this analysis, a variant of the English EPP[+DP] feature is on T˚, which targets weak pronominals, raising them to SpecTP. On the surface, agent pronouns precede the verb but follow the TAM marker, meaning the TAM marker must be positioned higher up in the syntax than SpecTP, where the agent pronoun is.

(5)
Samoan (Collins 2017: 32) 'Ole'ā 'ou lē alu. fut 1sg neg go 'I will not go.' Under a T-to-C analysis, TAM raises out of TP, meaning the subject pronoun in SpecTP will follow the TAM particle in the surface order. Without T-to-C movement, TAM remains in T˚, while the subject pronoun raises to SpecTP, which would mean the subject would precede the TAM particle, rather than follow it. T-to-C movement therefore provides an explanation for the TAM-initial word order in Samoan.
Note that all of these previous accounts assume a simple clause periphery, with the C head taking TP as a complement, which is what this squib will call into question. I will propose that a single CP cannot account for the data seen in these languages. 5 It has been also argued for Niuean that TAM particles share properties with complementisers (Massam 2000). In operator-extraction contexts, the clause-initial particle encodes both the tense of the clause, and also clause type, as either a main clause or a subordinate clause (Seiter 1980). For example, in main clauses the TAM marker which expresses future tense is to, and the particle which expresses the same tense in relative clauses is ka. As these particles have combined the grammatical information of both complementisers and TAM markers, Massam (2000) posits that the C and T heads are a morphologically merged single particle. Note, however, a contrasting account exists in Clemens (2014). Clemens takes a head-movement analysis of Niuean (contra Massam), arguing via prosody that T-to-C movement occurs in this language.

6
Note, however, this is not the only analysis of the complementary distribution of C and TAM markers, and that there can be an alternative explanation. Mosel & Hovdhaugen (1992: 595) state that the complement clauses with ona are placed under the scope of the TAM particle in the matrix clause. This has several consequences. Firstly, in multi-clause sentences with a complementiser ona, TAM particles are sentential modifiers, which can provide TAM information for the entire sentence. As a result of this, the complementiser ona and a TAM particle will never appear together, as there is already a TAM particle in the sentence-initial position, which gives the sentence its TAM information. As the complementiser and TAM will never overtly co-occur, this gives the appearance of a complementary distribution -where one exists, the other cannot. In this analysis, the lack of co-occurrence may simply relate to the lack of a need for a TAM to be present in the subordinate clause; the apparently complementary distribution could be simply a by-product of the matrix TAM particle controlling tense for the entire sentence. However, as an anonymous reviewer pointed out, this would be highly unusual, as it is generally understood that a CP selects a TP. As such, it would be unexplainable how a complementiser such as ona, in CP, could select a TP without a TAM particle being generated in T˚. Dispensing with the TP node altogether is the only way this analysis could play out, and this idea has its limitations as higher heads cannot dispense with lower nodes via selection.

Tokelauan
In each language above, one critical piece of evidence for T-to-C movement is that TAM particles and complementisers are in complementary distribution with one another, and as such will never co-occur. According to Custis (2004: 122) and Collins (2017: 30) (though not Otsuka), TAM only raises to C˚ when C˚ is unfilled. However, when C˚ contains an overt complementiser, T-to-C movement is blocked, and a TAM particle cannot be overt. Here I will present novel data from Tokelauan, another verb-initial Polynesian language, which illustrates that TAM particles and complementisers can co-occur.
Tokelauan has a complementiser particle pe, which introduces embedded indirect question clauses. Pe appears before the complement of a verb of knowledge, and translates roughly as 'if' or 'whether'. This complementiser can introduce verbal predicates, as well as nominal predicates (Hooper 1993: 277). As illustrated below, pe can occur with a subordinate finite clause, where a TAM marker is present within the embedded clause.
(6) Tokelauan Ko John na fehili mai pe na tunu e ai te ika. top John pst ask dir comp pst cook erg who def fish 'John asked who cooked the fish.' The ability for both the complementiser and TAM marker to coexist within one clause demonstrates they are not in complementary distribution. Furthermore, it can be shown that C and T are not simply combined as a syntactic complex head, after undergoing T-to-C movement. If the TAM marker and complementiser are somehow adjoined, it should not be possible to split them with a phrase. In (7) and (8) the complementiser and TAM particle are separated by a ko-topicalised phrase. This is a maximal projection, which therefore cannot be argued to be a clitic attaching within a morphologically merged C˚ and T˚. Tokelauan (Hooper 1993: 182) … fehili mai ki a te au pe ko au na hau i fea. … ask dir to art 1sg comp top 1sg pst come loc where. '… he asked me where I had come from.' This is not an isolated example. Examining the literature, we come across several cases of complementisers and TAM co-occurring, with phrasal units in-between. In (9), the complementiser kāfai 'if' is followed by a ko-marked topic and then TAM, while in (10) and (11) the complementiser is followed by an adverb and then the TAM particle.

(9)
Tokelauan (Hooper 1993: 127) Kāfai ko nā atu mua e fai tamate pe fai tiaki, kua comp top def skipjack first prs do lose-overboard or do foul-hooked prf fakahētonuga loa te taumanu i te taimi tēnā. be-heedless at-once def feeding-school loc def time dem 'If the first skipjacks are handled so that they fall back in sea, or get foul-hooked, … the feeding-school is heedlessly ruined at once at that time.' (10) Tokelauan (Hooper 1993: 127) Kae kāfai foki e hē lava tau meakai. ka lahi lele nā conj comp too PRS neg sufficient 2sg.poss food fut many int def mea ka tutupu mai kia te koe. thing fut happen dir to def 2sg 'But if your food is insufficient, there will be many things that will happen to you.' Tokelauan (Hooper 1993: 166) Kāfai foki e i luga ni ika i tō vaka comp too prs loc above def fish loc 2sg.poss canoe 'If there are fish on your canoe.' This data confirms that TAM cannot raise to C˚ if C˚ is the only complementiser position, as TAM can co-occur with complementisers. It is also worth noting that the word orders which prompted Collins (2017) and Otsuka (2005) to propose T-to-C movement also occur in Tokelauan. The basic word order of Tokelauan is VSO, but VOS alternations are possible for some speakers, similar to Tongan: Additionally, like Samoan, an agent subject may be expressed as a pronoun, which intervenes between the TAM particle and the verb: Tokelauan (Hooper 1993: 62) Na ia veloa te ika. pst 3sg spear-cia def fish 'He speared the fish.' In the analyses of Tongan and Samoan by Otsuka (2005) and Collins (2017) respectively, the alternating VSO/VOS word order and the presence of pre-verbal subject pronouns was evidence of DP-movement to SpecTP, due to an EPP feature on T˚. Without making claims about the syntax of VSO/VOS and pre-verbal subject pronouns in Tokelauan, if either results in an argument situated in SpecTP, there is reason to suggest that TAM does raise from T˚, although the C domain requires expansion to rescue this movement, as will be argued for later in this squib.

Tongan and Samoan
Like Tokelauan, there is evidence that certain complementisers and TAM particles can coexist in both Tongan and Samoan, indicating that T-to-C movement needs revisiting in these languages as well.
Samoan has two complementisers that can occur with TAM particles. Like Tokelauan, verbs of cognition or perception can introduce interrogative subordinate clauses, with the complementiser po or pe.

(14)
Samoan ( The original translation in Mosel & Hovdhaugen (1992) appears to be erroneous. This has been corrected in this squib. Likewise, Tongan also has complementisers and TAM particles co-occurring.

(17)
Tongan (Churchward 1953: 50) 'Oku 'ikai te u 'ilo pe 'oku lelei pe kovi. prs neg fut 1sg know comp prs good or bad 'I do not know whether it is good or bad.' As was seen in Tokelauan, it is possible to split C and TAM markers in Tongan with other morphological material, as seen below with a negative particle. 8 (18) Tongan (Broschart 1999: 107) Kuo pau ke 'oua te u 'alu. pfv necessary comp neg fut 1sg go 'I must not go.' (lit. 'It is necessary that I will not go.') To rule out the possibility that the negative particle could be a clitic (and thus creating a complex head C-NEG-T), and to provide support that T and C are indeed independent in Tongan, consider the following argument. Tongan exhibits a noteworthy conditional structure, the ka-ka construction, which involves a pronominal clitic intervening between a complementiser and a (phonologically identical) TAM particle (Ball 2008): (19) Tongan (Churchward 1953: 42) Ka nan ka fehu'i mai, te u tala kiate kinautolu. if 3pl fut ask towards fut 1sg tell dat 3pl 'If they ask me, I will tell them.' (20) Tongan (Dukes 2001: 76) ka ke ka ha'u, 'oku 'atā 'a to'ohema if 2sg fut leave prog free abs left 'if you leave, the left will be free.' In this construction, the complementiser is able to host a subject clitic (Ball 2008: 157), which indicates the complementiser is a freestanding (unbound) morpheme. Crucially, the TAM particle in Tongan also has the ability to host subject clitics, meaning it too is a morphologically independent word. Tongan subject clitics also surface between the TAM particle and the verb (Otsuka 2005). On the basis of prosody, subject clitics are shown by Custis (2004) to attach to the TAM, as there is a stress shift when the clitic appears between the TAM marker and the verb. This stress pattern clarifies that the clitic has attached to the TAM marker, which is only possible if the TAM marker is a freestanding morpheme. Note it is not possible that TAM is itself a clitic which combines with the subject clitic/complementiser to form an independent word, as TAM can be found without a complementiser or subject clitic, as in (1). In this case the TAM particle would have no host to attach to. In (19) and (20) there is clear evidence that the complementiser is a freestanding morpheme, while in (21) there is support that the TAM is also a free morpheme. Combining these two pieces of information means that when both the complementiser and TAM marker are observed co-occurring, we would have to assume that C˚ and T˚ are independent heads in Tongan, which provides evidence against a [T + C] complex head.
To sum up, this section has provided evidence that complementisers and TAM particles can cooccur, and that they are each a separate morphological entity that exists independently from each other. This is supported by examples where phrasal constituents can separate the two, which rules out the possibility that T˚ and C˚ form a complex morphological head. Furthermore, in Tongan, it was seen that both T˚ and C˚ can host clitics, which further supports an analysis that they are independent particles in the language.

Proposal
From the data above, I have argued that the current T-to-C movement account must be adjusted in order to derive the patterns seen. In this section I suggest that we can best understand the data and maintain the general essence of head-movement in the earlier works if we adopt an expanded left-periphery analysis (Rizzi 1997) for these languages, which I illustrate with a specific proposal for the heads in Tokelauan. This squib proposes that TAM particles in Tokelauan raise beyond T˚ to the left periphery, but to a lower head than the one which hosts complementisers pe and kāfai. The details of this proposal are discussed below.
I follow Rizzi (1997) and argue for an expanded left periphery in Tokelauan, building on similar approaches by Pearce (1999;to appear) and Massam (2020) for Māori and Niuean respectively. Taking this approach, I argue that there are multiple functional projections within the C-domain. Highest is ForceP, where I suggest complementisers which take indicative complements, such as pe and kāfai, must reside. Below is a TopicP, where ko-marked DPs raise to, while FinP is the lowest. FinP is where complementisers which take non-indicative complements, such as ke and oi, are situated. I argue that when TAM undergoes head-movement, it targets the Fin˚, and thus is predicted to be in complementary distribution with ke and oi. TAM, in this analysis, is (correctly) predicted to co-occur with the higher complementisers pe and kāfai. The data and the proposal are discussed in more detail below.
Firstly, in Tokelauan there are clear cases where a complementiser is in complementary distribution with TAM, and as such they cannot co-occur. This is seen for ke and oi, two complementisers which under no circumstance can exist with a TAM particle. I continue to advance the proposal discussed for Tongan and Samoan that when a complementiser and TAM are in complementary distribution, we can claim that the TAM has raised to that complementiser's position, and as such, they will never co-occur. This being said, this squib proposes that TAM raises to the complementiser position of ke/oi.
While there are complementisers which do not co-occur with TAM, there are other complementisers in Tokelauan that do, such as pe and kāfai. I claim that the ability to co-occur with TAM demonstrates they reside in a different functional projection from ke and oi. This position must be higher than that final resting place of TAM, as pe/kāfai always precede a TAM marker. Therefore, the functional projection of these complementisers is higher than the projection of ke/oi, where TAM when raised in ordinary sentences.
This proposal is supported by the distribution of ko-marked nominals. As mentioned before, Tokelauan has a process of ko-topicalization, where a nominal is fronted to the left periphery, and marked with ko. 9 As shown in (7), (8) and (9), ko-marked nominals follow pe/kāfai, which means the Topic projection is lower than the pe projection.
However, in ke sentences, when a nominal from the subordinate clause is ko-topicalised it appears preceding the complementiser (see 24 and 25). Assuming the Topic position is a static landmark in all sentences, this data demonstrates that the ke/oi position is below the Topic projection.10 The result of this data is a clear understanding of the makeup of the left periphery in Tokelauan. I will follow Rizzi's original proposal in terms of the names of the functional projections, using ForceP and FinP. In the highest projection pe and kāfai are found, followed by the Topic projection. Lastly, the complementisers ke and oi are found in FinP. It is noteworthy that complementisers ke/oi introduce clauses that translate as infinitives/subjunctives in other languages, and as such, naturally fit into the Fin projection. As ke/oi are in complementary distribution with TAM, we assume that TAM raises to Fin˚ from T˚. This analysis accounts for all the data in this squib, and gives a clearer picture of the left periphery in Tokelauan. 9 Note that this process is called ko-topicalization in the literature, so this squib uses the same term, although it is possible that these nominals are focused (Brown & Koch 2016), rather than topicalised. Crucially, all ko-marked DPs appear in this position, higher than ke and oi, and lower pe and kāfai.
10 Note there is a TAM particle which had an identical phonological form as the complementiser ke. The modal particle ke has a "desiderative force", used in imperative sentences, according to Hooper (1993:54). There are several reasons against claiming that the sentences in (24) and (25) have no overt complementiser and the ke particle is simply a TAM particle. Firstly, there is clear evidence that there is a complementiser ke, as this patterns together with oi, and they are occasionally interchangeable, as seen in (i) (see Hooper 1993:280 for why this occurs).
(i) E mafai oi/ke i kinei ia Rangi. tam possible comp loc here abs Rangi 'It is possible that Rangi is here.' A second reason is that if ke was a TAM in (24), it should be possible to replace it with a different TAM. As shown in (ii), replacement with a past tense TAM is not grammatical.
(ii) *Na taumafai ia John na hao te vaka mai te afā. pst try abs John pst escape def boat from def hurricane Intended: 'John tried to escape the ship from the hurricane.'

Conclusion
This squib has examined arguments for T-to-C movement in Tongan and Samoan, to determine whether this process can be extended to another Polynesian language, Tokelauan. On the basis of C and TAM co-occurrence in these languages, this squib argued that T-to-C movement needs to be refined to explain the TAM-initial word order. The squib has suggested an expanded left periphery for Tokelauan, where different complementisers are situated in two positions: higher complementisers in ForceP and lower complementisers in FinP. I have argued that TAM raises to the lower complementiser head Fin˚, which accounts for the complementary distribution of TAM and these complementisers ke and oi. The higher complementiser head Force˚ contains pe and kāfai, those which can co-occur with TAM. The distribution of ko-marked topics further supports this structure, as these topics follow ForceP complementisers, and precede FinP complementisers.

Abbreviations
Some original abbreviations have been altered to follow the Leipzig Glossing Rules where these apply. Additional abbreviations used in glosses: dir=directional particle; emph=emphatic; tam=tense/aspect/modal particle; cia=agentive verbal suffix.