To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   The Final-over-Final Condition and Particles Theresa Biberauer (Note: All acknowledgements relating to this paper appear in the general acknowledgements at the start of the currently in press MIT volume. If you’ve ever discussed (FOFC and) particles with me, thank you again for having been willing to teach me about these peculiar elements. As my work on FOFC and particles is on-going, I remain keen to keep discussing specific cases and learning about those and the phenomenon more generally – so do please contact me if you have comments, questions, and/or other input. Also note: I’m aware that the References at the end of this (immense) chapter do not in every respect match up to what is cited in the main text. This has been addressed in the final version. Do feel free to contact me for clarifications relating to referencing – and other things – in this version, though. And note most particularly: As of January 2017, the constraint formerly known as the Final- over-Final Constraint is, officially, the Final-over-Final Condition. Logically …) of this talk 1. Introduction   This chapter focuses on a diverse range of structures, all containing elements that have, in one context or another, been referred to as particles. As already noted in chapter 2, particle- containing structures superficially seem to be able to violate the Final-over-Final Condition (FOFC) rather readily. While V-O-Aux structures containing inflecting auxiliaries of the kind found in Germanic, Hungarian, Basque, and many other languages appear to be ruled out, V- O-Aux structures where Aux does not inflect do not seem to be subject to the same condition. The examples in (1) illustrate: (1) a. Tā chī-le fan le. [Mandarin] 3SG eat-PERF food PERF ‘He has eaten.’ (Paul 2014:86)1 b. Ego psis dio avga iton. [Cappadocian Greek] 1SG bake.1SG.PERF two eggs PAST (= 3SG.IMPERF.BE) ‘I had baked two eggs.’ (Español-Echevarría 1994:1) c. Bɨs sa ja tebɨre ga. [Bagirmi] dog eat meat yesterday COMPL ‘The dog has eaten the meat.’ (Stevenson 1969:85) That noninflecting auxiliaries do not behave like their inflecting counterparts has been well- known since Greenberg’s groundbreaking typological work: in establishing his universals, he systematically excluded “uninflected auxiliaries,” given the fact that they so clearly do not pattern like inflected ones (see Greenberg 1963:85, 93). Similarly, Dryer (1992:99) pinpointed the ability to “bear all or some of the verbal inflections associated with the clause”                                                                                                                 1 Paul (2014) glosses clause-final le as C , reflecting her analysis of this element. The gloss given Low here instead prefigures the analysis to be presented in section 4.4.2, in terms of which this le and verbal le share certain meaning components, with clause-final le being a vP-internal element (see also To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   as a difference between his categories of “auxiliary verb” and “tense/aspect particle,” concluding that the former should therefore be regarded as heads (“verb-patterners”) and the latter as modifiers (“object-patterners”). That the presence vs. absence of inflection is a relevant consideration in establishing why structures like (1a–c) are possible, whereas counterparts featuring inflected auxiliaries are not, is also strongly suggested by language- internal contrasts of the kind illustrated for Cappadocian Greek and Bwe-Karen below. Let us first consider Cappadocian Greek: (2) a. Ego iha psisi dio avga. [Standard Modern Greek] I have.1SG bake.PTCP two eggs ‘I had baked two eggs.’ (Español-Echevarría 1994:1) b. πῆγα ἐτόν.2 [Cappadocian Greek] ˈpiɣa eˈton go.PAST.1SG PAST (= 3SG.IMPERF.BE) ‘I had gone.’ (Krinopoulos 1889:37) c. νίφτα ἦτον. ˈnifta ˈiton be.washed.PAST.1SG PAST (= 3SG.IMPERF.BE) ‘I had been washed.’ (Archelaos 1899:141) d. πῆγαν ’τόν. ˈpiɣan don go.PAST.3PL PAST (= 3SG.IMPERF.BE) ‘They had gone.’ (Alektoridis 1883:489) As the contrast between Standard Modern (2a) and Cappadocian (2b–d) Greek shows, there is more than just a positional difference between the auxiliaries in these two varieties. Standard Modern Greek’s preverbal auxiliaries exhibit the kind of finite inflection familiar from well- studied Western European languages: in compound tenses, the auxiliary exhibits agreement and tense inflection, while lower verbs surface without this inflection, instead being marked as nonfinite in some way. By contrast, Cappadocian Greek’s final auxiliaries systematically fail to inflect: they are consistently third-person singular imperfect past forms, with the lexical verb instead bearing the full tense-aspect and agreement morphology that one would expect to find on a finite verb; as the examples above show, there can be both aspect and agreement discrepancies between the fully inflected lexical verb and the invariant clause-final auxiliary. This is a property of Pontic Greek varieties more generally (see Neocleous and Sitaridou in preparation). In these varieties, then, it is always a lexical verb that bears full finite inflection, with the BE-derived auxiliary interacting with neither  the φ-/aspectual nor the argument structure (cf. (1b) and (2a)) properties of the clause. Let us now consider Bwe-Karen (Tibeto-Burman):                                                                                                                 2 Thanks to Petros Karatsereas, Nicos Neocleous and Ioanna Sitaridou for discussion of the Cappadocian and, more generally, Pontic Greek data and for pointing me to the examples given in (2b–d). The variation in the realization of the final element is phonologically driven and thus left aside here. To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   (3) a. ce-ɗɔ mi jəә-khɔ phi má nɔ (*jəә-khɔ)? [Bwe-Karen] 3- say C 3- FUT take what 3- FUT ‘What did he say that he would take?’ (Henderson 1997:187) b. yəә- ca (*lɔ) dɛyo lɔ. 1SG-see ASP picture ASP ‘I am looking at a picture.’ (Henderson 1997:39) As in the case of Cappadocian Greek, we see a clear distinction here between the placement of inflected and the placement of uninflected elements. As (3) shows, Bwe-Karen features clause-final aspectual markers that do not inflect and that must necessarily surface in clause- final position. In this case, however, it is important to note, as Philip (2012:115) points out, that the appearance of inflection does not actually point to the existence of inflected auxiliaries in Bwe-Karen; instead, the “agreement” in (3a) is a proclitic pronoun, which is therefore systematically absent in structures containing a full DP subject (see Swanson 2011:24 for an overview of the Bwe pronoun system). Nevertheless, the contrast between (3a) and (3b) is instructive in the wider context, as Bwe-Karen’s final auxiliary elements differ strikingly from its numerous preverbal auxiliaries and also from its lexical verbs in being unable to host proclitic pronouns (see Swanson 2011 for detailed descriptive discussion of the Bwe verbal system). This renders these elements exceptional within the verbal system, as all other verbal elements—auxiliary, serial-verb, and lexical—can combine with person marking where they are the hierarchically highest verb in the structure. Acquirers, then, will have a very clear person-marking cue, parallel to that which acquirers of inflecting languages get, as to the formally quite distinct status of clause-final verbal elements. As will become clear in section 4, this cue is reinforced by numerous further indications in other parts of the grammar that a distinction needs to be drawn between heads that contribute directly to the extended projection (i.e., FOFC-relevant heads) and semantically related elements that do not (i.e., elements that do not “count” for FOFC purposes) (regarding extended projections, see Grimshaw 1991 et seq.). For the moment, I introduce just one of the latter, a recurring phenomenon in predominantly head-initial languages with an inventory of (apparently FOFC- violating) head-final functional elements: (4) a. θrɪɗokha ɗó nu ɠɔ̀ yà θí lò cɛ (bé).3 [Bwe-Karen] rhinoceros animal.CLF DEM PROB step.on die finished 3SG IRR ‘The rhinoceros might trample them to death.’ (Swanson 2011:42) b. tʰan3 saa1maat3 tʰam (daj3). [Thai] 3SG able do able ‘He can do it.’                                                                                                                 3 The optionality-signaling brackets have been added to Swanson’s (2011) original example here. Swanson does not explicitly mark optional elements, but the discussion of irrealis/epistemic markers makes it clear that the obligatory element is the preverbal one, many bé-less examples being cited in the course of this discussion. To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   (Hanna 2010:17) c. too1daj1 baw2 ʔaat2 cak2 kææ3 (daj3). [Lue] anybody not able IRR solve able ‘Nobody could solve (the riddle).’ (Hanna 2010:16) In all of these examples, the element highlighted in grammatical descriptions as the auxiliary verb—irrealis ɠɔ̀ ‘might’ in (4a), and ability saa1maat3 and ʔaat2 in (4b,c)—may cooccur with a clause-final element associated with the same interpretation. In all cases, the final but not the initial element is optional, clearly signaling that the preverbal element is the “real” auxiliary (head).4 What we have established so far, then, is that there may be good reasons to think that auxiliary particles are formally distinct from the type of auxiliaries found in FOFC-respecting languages. What I will aim to achieve in the following sections is, first, to show that particles more generally do not constitute a threat to the universality of FOFC, interpreted (i) as a hierarchical universal (i.e., as a constraint on permissible narrow-syntax/narrow-syntax- internal phrase structure configurations; see Whitman 2008), and (ii) as a constraint that is relativized to extended projections in the manner stated in (5) (pace Abels 2013, Hawkins 2013, Sheehan 2013a, this volume, Whitman 2013, Etxepare and Haddican 2014, Erlewine to appear a,b, Zeijlstra 2015, Richards 2016; see chapter 1 for overview discussion): (5) A head-final phrase αP cannot dominate a head-initial phrase βP where α and β are heads in the same extended projection. (cf. Biberauer, Holmberg, and Roberts (BHR) 2014) Second, I will aim to explain why particles are such prolific apparent violators of this condition. In the course of this discussion, I will consider a wide range of particle types and demonstrate that there are various formal structures that would, on a narrow-syntax-internal, extended-projection- oriented interpretation of FOFC, be predicted not to produce a FOFC violation, all of which seem to be attested in the particle domain. This will lead directly to my third objective, which is to show how the FOFC-motivated investigation of particles reported here has led to what I believe to be a new discovery about the distribution of particle elements more generally: namely, that they are necessarily peripheral elements in a sense to be made precise. In other words, their distribution, like the distribution of disharmonic word order more generally, is more regular than has previously been thought. The rest of this chapter is structured as follows. Section 2 presents an overview of the types of apparently FOFC-violating structures that are found in the particle domain. Section 3 introduces the formal configurations that would give rise to superficially FOFC-violating structures without actually violating the version of this condition presented in (5). Section 4 shows that all of these configurations exist. Section 5 discusses the consequences of this fact and concludes.                                                                                                                 4 Interestingly, as Haas (1964:xxii) notes, Thai grammatical tradition refers to verbal words preceding the main verb as auxiliaries, while those following the main verb are called secondary verbs. The practice of distinguishing between positionally distinct elements with related aspect, tense, and mood semantics in languages of this type is therefore well- established. To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   2. Apparently FOFC-Violating Particles: The Empirical Facts The objective of this section is to illustrate the kinds of particle-containing constructions that have led to doubts concerning the universality of FOFC. 2.1. Final Auxiliary Particles We have already considered some examples of VO structures featuring a final auxiliary (see (1)–(4)). This pattern is particularly widespread in East Asian languages and it also surfaces, though to a lesser extent, in certain Central African languages (see Dryer 2009b:344–345).5 Worth noting in relation to the latter is that it always seems to be the case that V-O-Aux in languages that permit it is a minority pattern alongside Aux-V-O and/or inflectional tense/aspect/mood (TAM) marking (see Dryer 2015e). Consider the case of Bagirmi (Bongo- Bagirmi, Nilo-Saharan; Chad). This language features (prefixal) aspect marking on the lexical verb (6a,b), preverbal (agreeing) mood markers (6c), and a single6 clause-final completive particle, ga (6d) (and a clause-final question marker (6b); see also below): (6) a. ŋgab kä-pa kag(a). [Bagirmi] man IPFV-split wood ‘The man splits the wood.’ (Stevenson 1969:102) b. Q: boukar táɗ djùm tɛ́ŋ làbà sà ksàa wà? boukar PFV.do gruel millet or PFV.eat INF.eat Q ‘Did Boukar cook millet gruel or did he eat it?’ A: boukar táɗ djùm tɛ́ŋ táɗà. boukar PFV.do gruel millet INF.do ‘Boukar COOKED millet gruel.’ (Jacob 2010:129) c. (née) nəә́ ndugo kìtàb kɛɗɛ. 3SG 3SG.FUT IPVF.buy book INDEF ‘She/He will buy a book.’ (Jacob 2006:31) d. Bɨs sà ja tebɨre ga. dog eat meat yesterday COMPL ‘The dog ate the meat yesterday.’ (Stevenson 1969:85, cited in Dryer 2009b:344) As these examples show, there are clear formal distinctions between finite and nonfinite verbs                                                                                                                 5 The languages that Dryer (2009b:355) lists as having “VOAux order, for at least some Aux” are Bimobo (Gur, Niger-Congo), Adioukrou (Kwa, Niger-Congo), Mumuye, Linda (Adamawa-Ubangi, Niger-Congo), Birom (Platoid, Niger-Congo), Kresh (Kresh, Nilo-Saharan), Baka, Bongo, Jur Mödö, Ngambay, Mbaye, Bagirmi (Bongo-Bagirmi, Nilo-Saharan), Dholuo (Nilotic, Nilo-Saharan), Moru, Avokaya, Logbara, Ma’di (Moru-Madi, Nilo-Saharan), Kera (East Chadic, Afro-Asiatic), and Musgu (Biu-Mandara, Afro-Asiatic). 6 Peggy Jacob (pers. comm.). To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   in Bagirmi, with finite verbs systematically preceding the object (and nonfinite verbs), and post-object verb placement being possible only in verb-doubling constructions (6b) (see Jacob 2013 for discussion of the focus-backgrounding conventions that account for this placement). Completive ga, then, evidently does not occupy a position associated with either finite verbs or nonfocused verbs more generally. A similar pattern emerges in Ngambay (Bongo-Bagirmi, Nilo-Saharan; Chad, Central African Republic): (7) a. m- Îsi/ m-ár m-úsa da. [Ngambay] 1SG-PRES/1- PRES 1- eat meat ‘I am eating meat.’ (Heine and Reh 1984:126, Vandame 1963:94–96) b. m- Îsi/ m-ár mba k- ùsà da. 1SG-PRES/1- PRES for NOM-eat meat ‘I am eating meat.’ (Heine and Reh 1984:126, Vandame 1963:94) c. m- ā k- ào àl ngà. 1SG-FUT NOM-go NEG REPEATED ‘I will not go again.’ (Vandame 1963:118, cited in Dryer 2009b:344) Here we see that Ngambay makes use of inflected auxiliaries that systematically occupy a preverbal position.7 Strikingly, agreeing auxiliaries may either cooccur with agreeing lexical verbs (7a) or select for a nominalized complement, which, in the progressive, is introduced by a preposition (7b); k- marks the infinitive in (7b,c).8 The formal and distributional differences between the finite and nonfinite Ngambay verbal forms and clause-final aspect-marking ngà are thus again evident. And similar observations can be made about Mbaye, another of the VO Bongo-Bagirmi languages that Dryer (2009b:344) identifies as featuring at least one clause- final auxiliary (see note 5; see Keegan 1997 and Anderson 2011 for discussion). Three final Central African V-O-Aux languages highlighted by Dryer (2009b) that I will comment on here are Mumuye (Adamawa-Ubangi, Niger-Congo; Nigeria), Dholuo (Nilotic, Nilo-Saharan), and Ma’di (Moru-Madi, Nilo-Saharan). Dryer (2009b:356) presents the following data showing that Mumuye has both clause-final aspect (8a) and mood (8b) particles:                                                                                                                 7 That the auxiliaries in (7) are in fact inflected, rather than associated with clitic pronouns as in the Bwe-Karen case discussed in section 1, is clear from the discussion in Ndjerareou, Melick, and Moeller 2010: there are clear discrepancies between the realization of agreement (which is, for example, suspended in the second and third person singular) and the availability of clitic pronouns for all persons. 8 That the inflected tense-marking elements in (7b) are in fact auxiliaries and not copular elements is clear when one considers their overall distribution, and also when one considers the form of equative structures, which do contain a copula: (i) ku˧nja˥ to˧ yel˩˨. [Ngambay] chicken 3SG.COP bird ‘A chicken is a bird.’ (Here, ˧ is a mid tone, ˥ a high tone, and ˩˨ a low to mid tone; see Ndjerareou, Melick, and Moeller 2010:38.) To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   (8) a. znàso̘ baasé ranti yé. [Mumuye] Znaso mimic Ranti PERF ‘Znaso has mimicked Ranti.’ (Shimizu 1983:107) b. znàso̘ dé baasé ranti ni. Znaso PERF mimic Ranti IMMED.FUT ‘Znaso is about to mimic Ranti.’ (Shimizu 1983:112) The discussion in Krüsi 1978 further reveals that this language has numerous other final particles, expressing meanings relating to aspect, mood, negation, interrogativity, and discourse-connectedness. Particularly important for our purposes is, first, the fact that Mumuye again appears to have auxiliary(-like) elements that surface in the clause-medial position that one would expect for a VO language, dé in (8b) being a case in point (see also na in (9a) and in Krüsi’s illustrative narrative text). Second, closer consideration of the elements discussed in Krüsi 1978 shows (i) that the elements occurring in final position seem to be quite numerous and (ii) that at least some final elements appear to be multifunctional, having considerable positional freedom. A notable case in point is ne, which serves as a (partly discourse-oriented) continuative marker (9a), coordinator (9b), and general linker (9b): (9) a. ɓayeh na wa’n ne. [Mumuye] Bayeh PAST sit CONTINUATIVE ‘Bayeh was sitting.’ (Krüsi 1978:271) b. ne tó ne wu ti ɓeehsan do yu. LINK say LINK he fix friend still hair ‘And he said that he was still fixing his friend’s hair.’ (Krüsi 1978:272) Patterns of this sort are also very evident in both Dholuo and Ma’di, much-better-studied languages on Dryer’s V-O-Aux list. Descriptions of Dholuo consistently refer to medial TAM particles, surfacing in preverbal position, in front of a lexical verb that may, under the relevant circumstances, inflect for both subject and object agreement and aspect (see, e.g., Omondi 1982, Tucker 1994, Ojwang’ 2008, Cable 2012). Strikingly, the TAM markings employed in Dholuo-English code-switching are consistently and exclusively of the preverbal type (Ochola 2006), suggesting that these are the core TAM elements in the system (see also notes 4 and 9, and the discussion of primary and secondary TAM marking below). This impression is reinforced by the observations in the literature querying the formal status of Dholuo TAM elements more generally: Cable (2012:657n10), for example, notes that the preverbal TAM elements form a closed class, with meanings reflecting those of the (inflectional) tense paradigms of surrounding languages (e.g., Kikuyu; Schwartz 2003);9 following Omondi                                                                                                                 9 Cable (2012:656) lists the preverbal tense particles in (i): (i) a. a(ye) very recent past (just happened) b. ne(nde) recent past 1 (any time today) c. nyo(ro) recent past 2 (any time yesterday) d. nyo(cha) recent past 3 (any time more than two days ago) e. ne(ne) remote past tense 1 (at least several days ago) f. yande remote past tense 2 (at least several days ago) To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   (1982) and Tucker (1994), among others, he raises the possibility that some of these particles may in fact be adverbs, citing their apparent positional freedom in support of this idea. (Omondi and Tucker note that many of the medial T-elements are homophonous with postverbal adverbs; see also Cable’s note 10. As we will see in later sections, this type of apparent homophony, which could equally be multifunctionality involving underspecified elements (see, e.g., Biberauer 2011 et seq., Duffield 2013a, 2014a,b, Wiltschko 2014), is a recurring theme in the particle domain, not only in Central Africa but also more generally.) Blackings and Fabb’s (2003:chap. 18) discussion of adverbial placement possibilities in Ma’di shows clearly why an adverbial analysis needs to be carefully considered where elements expressing TAM-related semantics surface clause-finally, and also why this is particularly important when dealing with languages that do not overtly mark TAM on lexical verbs. Ma’di features both free and fixed-position adverbial elements. The example in (10) illustrates the placement options available to free adverbials, many of which express temporal and also discourse-related meanings (see also sections 2.2–2.4;10 @ in each case marks a possible placement option); (11) gives some examples of adverbials that obligatorily surface clause-finally: (10) dr!̀ádrū ‘now’ [Ma’di] @ má @ lɛ̀ @ mū-lé @ ēɓù gá @ 1SG want N- go.SB work LOC11 ‘Now I want to go to work.’ (Blackings and Fabb 2003:479) (11) a. ɔ́p! ́oɗū ízí gbù. [Ma’di] Opi 3.take woman before ‘Opi has married before.’ (Blackings and Fabb 2003:509) b. dʒì mī l!́/ *gbírí close eye completely/completely ‘shut eye completely/tightly’ c. kɔ̄ ōɗú l!́/ *gbírí fall asleep completely/completely ‘fall fast asleep’                                                                                                                                                                                                                                                                                                                               g. ang’ near future tense 1 (later today) h. kiny near future tense 2 (some time tomorrow) i. orucha near future tense 3 (at least two days from now) Of these, a small subset (e.g., a(ye), ang’) cannot occur clause-finally. Importantly, many postverbal TAM-related forms cannot surface in the preverbal position occupied by TAM elements. There is therefore a partial overlap between medial and final elements, with most of the former being able to surface in final position, but many of the latter not being able to surface preverbally. 10 See Blackings and Fabb 2003:476 for discussion of constituents within which these adverbials may not surface. Essentially, these appear to be what we might think of in Givónian terms as satellite- internal positions; that is, adverbials may not surface internally to DPs and PPs. 11 Here, N signifies ‘nonpast’ and SB that the associated verb is a subordinate form. See section 2.2 for further discussion of Ma’di verb forms. To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   d. dʒì tī gbírí/ *l!́ close mouth completely/completely ‘close mouth completely’ (Blackings and Fabb 2003:504) The absence of overt TAM marking on the verb combined with the presence of obligatorily final TAM-related elements12 readily creates the impression that there may be clause-final TAM positions in Ma’di.13 Importantly, though, closer investigation reveals the semantically specific nature of these final elements. Consider, for example, the lexical restrictions on the use of the completives in (11b–d), which are representative of a more general pattern: Ma’di completives are numerous and have highly specific lexical requirements (Blackings and Fabb 2003:504–505). Taking into account what is now known about grammaticalization (see, e.g., Heine and Kuteva 2002, Hopper and Traugott 2003, Roberts and Roussou 2003, van Gelderen 2004, Roberts 2007b, and Narrog and Heine 2011 for overviews), it is clear that Ma’di’s completives do not have the semantic profile typically associated with functional heads (Asp, T, M, etc.). They also do not have the formal profile of such heads, being amenable to “nominalization with r!̀ [glossed DEF]” and also, in this case, to left-dislocation (Blackings and Fabb 2003:499). This is a property they share with other final adverbials, and with modals and negative elements, to which we will return in section 2.2. Crucially, this is very different from what we see in the context of languages in which auxiliaries cannot occur in V-O-Aux configurations: in these languages, auxiliaries resist independent fronting (e.g., topicalization and/or focalization operations) or, in systems permitting these operations, Stylistic Fronting (Holmberg 2000b, 2005) and predicate doubling (see, e.g., Güldemann 2010, Biberauer 2013). The completive and, more generally, auxiliary elements that surface finally in Ma’di appear to differ from FOFC-respecting auxiliaries, then, not only morphologically, by virtue of being consistently uninflected, but also in semantic and syntactic terms. Looking beyond Central African languages, patterns strikingly similar to those highlighted above repeatedly emerge. First, East Asian VO languages, for example, feature both initial and final auxiliaries (12). Second, in these languages a range of elements at different points along the grammaticalization spectrum and with varying s-selection requirements are available to express the meanings associated with some of these elements (13)–(14). Third, many of the apparently FOFC-violating auxiliary particles are superficially homophonous with elements able to surface in other positions (15)–(16): (12) a. Zhāng Sān néng qù Táiběi le. [Mandarin] Zhang San can go Taipei PERF ‘Zhang San can go to Taipei.’ [✓le > ABLE TO, *ABLE TO > le] b. Zhāng Sān kěnéng qù Táiběi le. Zhang San may go Taipei PERF ‘Zhang San may have gone to Taipei.’ [*le > MAY, ✓MAY > le] (Erlewine to appear a:9)                                                                                                                 12 Ma’di also has an extensive inventory of temporal nouns whose default position is clause-final (Blackings and Fabb 2003:522–532). 13 We will see in section 2.2 that Ma’di tense is in fact encoded via the morphological form of the lexical verb (and/or its interaction with other elements in the structure): so-called inflected verbs express nonpast tenses (present and future), while uninflected verbs mark the past (see Blackings and Fabb 2003:chaps. 7 and 8). To  appear  in:  M.  Sheehan,  T.  Biberauer,  A.  Holmberg  &  I.  Roberts  (eds.).  The  Final-­‐over-­‐ Final  Condition:  A  Syntactic  Universal.  Cambridge,  MA:  MIT  Press.   (13) a. se2- zo2 go2- fung1 seon3. [Cantonese]14 write-PERF DEM-CL letter ‘… have/has written that letter.’ b. se2- jyun4 go2- fung1 seon3. write-finish DEM-CL letter ‘… have/has finished writing that letter.’ c. se2- jyun4-zo2 go2-fung1 seon3. write-finish-PERF DEM-CL letter ‘… am/is/are done with writing that letter.’ (14) a. Tuō sā fēnzōng lìtóu huaĭ qiè [Yixing Chinese]15 he three minute in can eat le/ guāng/wuě sā ge pĭngguò. PERF/finish/ empty three CL apple ‘He can eat three apples to the core in three minutes.’ b. Tuō sā fēnzōng lìtóu huaĭ qiè guāng le/ wuě le he three minute in can eat finish PERF/empty PERF sā ge pĭngguò. three CL apple ‘He can eat three apples to the core in three minutes.’ (= (14a)) c. Tuō sā fēnzōng lìtóu xíng dăo/ *guāng/wuě he three minute in find arrive/finish/ empty éng ge pòngyòu. five CL friend ‘He found five friends in three minutes.’ d. Tuō sā fēnzōng lìtóu xíng dăo le éng ge pòngyòu. he three minute in find arrive PERF five CL friend ‘He found five friends in three minutes.’ (= (14c)) (15) a. Ông Quang được mua cái nhà. [Vietnamese] PRN Quang GET buy CL house ‘Quang was allowed to buy a house.’ Deontic (permission) b. Ông Quang mua được cái nhà. PRN Quang buy GET CL house ‘Quang was able to buy a house.’ Aspectual (accomplishment)                                                                                                                 14 Thanks to Chun Wai Leung, Joana Wat, and Cherry Lam for these data. It is worth noting that zo is the counterpart of postverbal le, and not of the clause-final le in (12). See section 4.4 for further discussion. 15 Thanks to Xuhui Hu for the Yixing data and for numerous discussions of the phenomena they illustrate.