Clause-final particles and focus in Eastern Cham

We demonstrate that clause-final particles in Eastern Cham (Austronesian: Vietnam) are right-branching syntactic heads that trigger predicate raising. This provides support for Simpson (2001)’s analysis of a clause-final modal found in Vietnamese, Thai, and other Southeast Asian languages, and militates against a mixed-headed analysis (pace Kayne 1994). Evidence for predicate raising comes from a novel diagnostic: the interaction between focus-driven object shift and multiple clause-final particles in one clause. Finally, we propose that clause-final particles are VP-level focus phrases, which divide a sentence into a focus and presupposition (cf. Rizzi 1997) and incidentally contribute modal or aspectual semantics.

1. Introduction.Eastern Cham (Austronesian: Vietnam), like many Southeast Asian languages, has a set of clause-final particles that typically mark mood or aspect, such as hu 'ROOT MODAL'.In unmarked contexts, these particles follow the verb and its arguments (1a), though they may be followed by discourse-level particles, like the polar question marker hlɛj.However, focussed material in the predicate can be extraposed to the right of these particles (1b).'I can eat mangoFOC.'Simpson (2001) analyses this phenomenon in a variety of Southeast Asian languages.Notably, numerous languages in the region have a clause-final modal likely related through borrowing or calquing (e.g.Thai dai, Vietnamese đượ c, Cantonese dak; cf.Enfield 2005).A natural surface analysis of (1a-b) would involve a mixed-headed approach (2a): whichever phrase is headed by hu is head-final (cf.Erlewine 2017 on Mandarin).However, these languages are all SVO and generally right-branching, which would make this phrase a conspicuous left-branching outlier.
Simpson proposes instead an analysis that preserves a right-branching structure by incoporating obligatory predicate raising (2b).The mixed-headed approach is more economical, involving fewer intermediate phrases.However, the predicate-raising approach keeps in line with the broader typology of these languages (cf.also Simpson 2005 on similar work explaining mixedheadedness in Southeast Asian DP's).
(2a) Mixed-headed derivation of (1a) (2b) Predicate-raising derivation for (1a) We propose that Eastern Cham clause-final particles must involve a right-branching, predicateraising structure, given evidence from the cooccurrence of multiple particles in one sentence.Eastern Cham has a set of particles which behave like hu (3a-b).
(3a) Clause-medial: c̥ ɛ̆ʔ 'near prospective', t̥ ɔʔ 'progressive', ka 'incompletive' (3b) Clause-final particles: o 'negative', hu 'root modal', v ̆ʔ 'iterative', mĭn 'emphatic', (p̥ lɔh) cɨ 'completive' In Section 2, we will show that the clause-final particles have rigid relative scope in the clausal spine.Evidence primarily comes from answers to polar questions, through the Predicator Test (Visonyanggoon 2000).In Section 3, it will be shown that focussed material in the predicate can be extraposed to the right of these clause-final particles.Then, in Section 4, the interaction between focus shift and multiple cooccurring clause-final particles will lend strong evidence that there must be predicate-raising in Eastern Cham.
2. Clause-final particles in the clausal spine.In this section, we will establish that Eastern Cham clause-final particles are syntactic heads with rigid relative syntactic scope.First, it is worth noting that these particles like hu are not verbal affixes.Words can intervene between the verb and hu (e.g. the object in 1a).These intervening words do not show obvious characteristics of noun incorporation, as they can be definites, whole phrases, and/or adjuncts and adverbials.Furthermore, Eastern Cham generally lacks bound morphology, along the same lines as Vietnamese.
The major evidence for this section, however, comes from answers to polar questions, via the 'predicator test'.The predicator test applies in languages where polar questions can be answered by a single verb or auxiliary.According to the test, only the structurally highest verb or auxiliary may be a felicitous answer.For example, in Thai, the main verb is a felicitous answer in the absence of auxiliaries (4b).But auxiliaries (5b), including clause-final auxiliaries (6b) outscope the main verb.
(4a) Thai (Visonyanggoon 2000: 117-118 The standard analysis of this phenomenon is that polar questions in these languages involve a polarity phrase head with a feature that selects for a verbal head (assuming a feature that targets both verbs and auxiliaries).Martins (1994) formalizes this as a ΣP, which dominates vP and whose head probes for the most local head with verbal features (citing López 1999).
This phenomenon is widespread in Southeast Asia.It is attested in Vietnamese (Austroasiatic), which is a language that has been in intense contact with Eastern Cham for centuries.Thompson (1987) describes "predicative fragments": modals and aspect markers that may answer polar questions (e.g.được 'root modal', rồi 'perfective').It is also attested in Indonesian (Austronesian), a language closely related to Eastern Cham.Sneddon, et al (2012: 324) list possible responses to polar questions, which are all modals or aspect markers (e.g.boleh 'may', sudah 'already').We thus have every reason to suspect that the predicator test should apply in Eastern Cham, and that modal auxiliaries should be pertinent verbal heads.
Indeed, the predicator test does apply in Eastern Cham.In particular, consider the examples in (7-10) with hu 'root modal' and v ̆ʔ 'iterative'.The main verb is a felicitous answer in the absence of auxliaries or serial verbs (7b).The iterative v ̆ʔ is a felicitous answer in (8b) as the highest auxiliary.However, the modal hu and the progressive t̥ ɔʔ are the only felicious answers in (9b,10b), as they outscope v ̆ʔ and the main verb.
We conclude that v ̆ʔ, hu, and t̥ ɔʔ are all syntactic heads with a verbal feature.Furthermore, we can assert the relative scope: hu 'root modal', t̥ ɔʔ 'progressive' >> v ̆ʔ 'iterative' >> main verb.Note that the root modal and progressive cannot cooccur, so their relative scope is unclear.
A secondary test to confirm this conclusion is fragment answers.In Eastern Cham, polar questions may also be answered by varying degrees of fragments, which we take to be a proxy for constituency.The felicitous responses in (11b,12b) are all constituents according to the scope above.In other words, v ̆ʔ 'iterative' is clearly lower in the clausal spine than hu 'root modal' and t̥ ɔʔ 'progressive'.In the vein of Simpson (2001), there are two obvious approaches to this data: a mixed-headed approach, where the clause-final particles are left-branching (13a), or a predicate-raising analysis (13b).To account for the relative order in (13b), we must posit a 'rolling-up' movement: first, the vP undergoes predicate raising to the specifier of the AspectP.Then, the whole AspectP undergoes raising to the specifier of the ModalP.(For Simpson, this movement targets unspecified intermediate YP's. ) While this roll-up movement may seem less preferable at first, the interaction between focus movement and multiple clause-final particles will militate in favor of this analysis.2(13a) Mixed-headed derivation of two (b) Predicate-raising derivation for two clause-final particles clause-final particles 3. Object shift and focus movement.In Eastern Cham, like some other languages of Southeast Asia, objects may sometimes be extraposed to the right of clause-final particles.Simpson (2001) reports that this occurs in Vietnamese, Middle Chinese, and Thai, and attributes object shift to focus.Object shift is also clearly related to focus in Eastern Cham: it is infelicitous for object shift to occur when the subject is focussed (14a), but it is preferred if the object is under focus (15a In the mixed-headed approach, a natural analysis of object shift would involve right-dislocation of foci, perhaps to a phase edge (cf.López 2009 on Romance clitic right-dislocation).Such an analysis is given in (24a).Under the predicate-raising approach, object shift involves focus movement to a FocP projected below the Modal/AspectP, then remnant movement of the vP as above (24b).Note that Simpson (2001) calls the ModalP "DeP" to associate it specifically with Thai dai and its areal cognates.Additionally, he treats predicate-raising as movement to an intermediate YP below TP.Here, instead, we treat predicate-raising as movement to the specifier of the Modal/AspectP in order to account for the data given in Section 4 with multiple clausefinal particles.Third, we assume that external arguments are generated low in vP, unlike Simpson, who treats them as selected for by the clause-final particle itself.
(24a) Mixed-headed derivation of object shift (b) Predicate-raising derivation of obj.shift The mixed-headed approach in (24a) generally accords with focus movement cross-linguistically (especially Romance clitic disclocation).The projection of a FocP between vP and the ModalP in (24b) seems to be more unusual, as focus movement generally targets phase boundaries.However, the evidence in Section 4 will strongly favor this analysis.This data presents significant problems for the mixed-headed approach.Given the syntactic scope laid out in Section 2 and the right-dislocation analysis in Section 3, the mixed-headed approach predicts that the (e-f) examples above should be the grammatical ones ( 27).The (a-d) permutations could be achieved through additional movement.However, it is unclear how the (e-f) examples would be blocked, making this analyis overgenerative.
(27) Mixed-headed derivation for *(26f) By contrast, the predicate-raising approach can easily accommodate for this paradigm.The (a) order results from raising the vP through Spec-AspectP to Spec-ModP (28a).The (b) order results from the roll-up style of movement described at the end of Section 2 (28b).Neither of these orders require any recourse to focus, as the object is non-focussed.
(28a) Predicate-raising analysis of (26a) (b) Predicate-raising analysis of (26b) When the object is under focus, the (c) order results from focus movement to a FocP projected above vP (28c), and the (d) order obtains when the object moves through the lower FocP to a FocP projected above AspectP (28d).
(28c) Predicate-raising analysis of (26c) (b) Predicate-raising analysis of (26d) Finally, the ungrammatical (e-f) orders are blocked for at least two potential reasons.First, Mod O and Asp O could have a feature such as [-Focus] or [+Given] that force their specifiers to contain non-focussed information.This analysis would require information structure features to percolate upwards through phrases (cf.Erteschik-Shir 2006).For the (e-f) orders to obtain, rollup movement of AspectP to Spec-ModalP must occur, but the AspectP necessarily contains a FocusP.In other words, it must contain focussed material, violating this featural restriction.
Second, these derivations could involve illicit movement.If predicate-raising and focus movement are both to be considered Ā-movement, this would violate a known constraint on remnant movement.According to Takano (2000) and others, there is a constraint that remnant movement of an XP and the movement out of that XP cannot be of the same type (e.g.*Āmovement/Ā-movement).The derivation in (28f) would violate this constraint.
(28f) Crashed derivation for (26f) Intuitively, this seems to be the right prediction.Object shift in Eastern Cham appears to a way to vacate focussed material out of the predicate before predicate-raising.The (e-f) orders both would involve putting the focussed material back into the phrase (AspectP) that undergoes raising, negating the intent of the previous movement operation.

Conclusion.
Overall, only a predicate-raising analysis in the vein of Simpson (2001) can account for the Eastern Cham clause-final particle and object shift data.This preserves a rightbranching structure in a language that is otherwise largely right-branching (pace Kayne 1994).
Based on data from multiple cooccurring clause-final particles, we propose slight modifications to Simpson (2001)'s analysis.Clause-final particles trigger predicate-raising to their own specifiers (unless one is willing to posit multiple intermediate YP's in between each Modal/AspectP).External arguments are generated low in vP, not by the clause-final particle itself.Third, we propose that there is a broader set of clause-final particles, occupying Modal/AspectP's.
It should be mentioned that this analysis of focus movement could be problematic for theories of information structure.Focus movement is generally treated as (a) movement either to phase boundaries (López 2009), (b) evacuative movement to avoid topic marking (Samek-Lodovici 2015), or (c) movement to a separate prosodic phrase due to prosodic constraints (e.g.Zubizarreta 1998).However, the focus movement in (28d) does not obviously meet any of these conditions.For (a), one would have to posit a phase boundary in between each modal or aspect marker, which is undesirable.For (b), the focus has already evacuated topic marking; why should it move again to the higher FocP?And for (c), the focus is not moved to a phrase edge; it is instead sandwiched between a modal and an aspect marker, with no attendant pause.More research is needed to ascertain the motivation for this focus movement.
A number of additional questions remain.First, how would this analysis derive focussed subjects?On this front, it should be noted that Eastern Cham is a topic-prominent language.
Focussed phrases generally must be clefted in order to be subjects, and unmarked subjects are usually given.Second, what are the properties of sentence-final, discourse-level particles, like the polar question marker hlɛj?After Erlewine (2017)'s work on Mandarin sentence-final particles, one may suspect that they would be less amenable to a right-branching structure.
Another question worth addressing is what are the interpretive differences between (26a-b) and (26c-d), respectively?The contextual control given in this paper is not fine-grained enough to detect a meaning difference between these pairs.Last, how do these facts compare to the contact language, Vietnamese?It is possible that this pattern is largely replicable with clause-final particles like nư ã 'more', đượ c 'root modal', and chı ... thôi 'only'.
As a final note, we will propose a general information structure for clause-final particles on analogy with Rizzi (1997)'s account of CP-level FocP's.Rizzi describes left-peripheral FocP's as partitioning the sentence into a focus (specifier) and a presupposition (complement), schematized in (29a).Clause-final particles perform a similar task under this analysis: they partition a predicate into a focus (complement) and a presupposition (specifier).
(29a) Generalized FocP, after Rizzi (1997) (b) Generalized clause-final particle phrase Both FocP and clause-final particles have additional, seemingly unrelated semantics.Leftperipheral FocP's also coincide with wh-movement.Despite the relation between wh-phrases and foci in standard alternative semantics, Rizzi and others consider this a coincidental overlap.Clause-final particles additionally add the semantics of modality or aspect, which could also be seen as a coincidental overlap brought about by a structural unity.
The two focus phrases also share optionality.If there is no focus movement, according to Rizzi, no FocP is projected.Likewise, if there is no focussed material in the predicate, no FocP is projected below clause-final particles.Clause-final particles, thus, may be the predicate-level analog to Rizzi's left-peripheral FocP.

4.
Multiple clause-final particles and object shift.Something that has received little attention in the clause-final particle literature is the interaction between multiple cooccurring clause-final particles.As shown in Section 2, it is possible for certain of these particles to cooccur, especially hu 'root modal' and v ̆ʔ 'iterative'.When these two clause-final particles cooccur, there are six possible permutations with the object, shown in (25).When the object is non-focussed (25a-b) are equally felicitous.When the object is focussed, (25c-d) are equally felicitous.The final two orders (25e-f) are ungrammatical.
(Roberts 1998)ht does not seem to determine object shift.In an out-of-the-blue context, objects with relative clauses (17a) or adjectives/numerals (17b) do not prefer to be shifted.Note that these sentences are felicitous if v ̆ʔ 'iterative' is clause-final.Syntactic weight may be a factor on some level, but it seems to be focus that primarily determines object shift.Next, while focus movement is claimed only to mark identificational/exhaustive focus (cf.Horváth 2010), this object shift does not seem to do so.A crucial test is association with 'only', which is a canonical identificational focus marker (É.Kiss 1998).In Eastern Cham, object shift is often accompanied by sĭt… mĭn 'only' (18a).However, it is easy to construct examples where 'only' occurs, but object shift is infelicitous (19a).Additionally, object shift is not felicitous with second-occurrence focus.Eastern Cham object shift targets discourse-new information.In particular, it only targets alternative sets which are unique to the immediate Question Under Discussion(Roberts 1998).Reconsider (18') below: object shift is felicitous, because the alternative set of foods is unique to the question.
#tŭʔ k̥ e kaw băŋ sĭt p ̥ ɔh ʔɔ̆ʔ mĭn hu However, in (19'), which fruit is a D-linked question, and the fruits are visible, so the alternative set of fruits is already in the shared context prior to the question.Accordingly, object shift is infelicitous.We claim, thus, that Eastern Cham object shift represents true new information focus.(19')CONTEXT:Whichfruit here can you sell me?[fruits are A: 'I can only sell you this papaya.'Itisworth mentioning here that 'object shift' is not limited to direct objects.Any focussed material in the predicate can be shifted, such as indirect objects (22) and locative adjuncts (23).EMPH 'You can only put beer in the kitchen.' This paradigm can be replicated with other pairs of clause-final particles, like hu 'root modal' and mı̆n/miː 'emphatic', and other focussed material in the predicate.(26)gives the same paradigm with hu and miː and the indirect object.Note that 'only that old man' is a secondoccurrence focus in (26c-d).Also note that scope is an outstanding issue in (26a-d): miː 'emphatic' seems to associate with 'only' in the subject, but it is unclear how the subject outscopes it.More research is needed to ascertain the relation between the emphatic and focus.