This is the Accepted Manuscript version of the following article:

Sarvasy, H. (2018). Multiple number systems in one language: split number in Nungon. Lingua, 201, 57-77, which has been published in final form at https://doi.org/10.1016/j.lingua.2017.09.001

This paper is made available in Western Sydney University ResearchDirect in accordance with publisher policies. Please cite the published version when available. Access to the published version may require a subscription.

© 2018. This manuscript version is made available under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-nd/4.0/

1 Multiple number systems in one language: split number in Nungon 2

3 Abstract

4 The grammars of some languages include two or more number subsystems that function in 5 different morphosyntactic contexts. These subsystems may vary in the number of distinct 6 number values distinguished, or in other ways such as obligatoriness. Such “mixed” systems 7 (Dixon 2012: 52) have been little-investigated. Corbett (2000: 89-94, 120–121) analyzes them 8 solely in terms of the nominal Hierarchy (Smith-Stark 1974, Silverstein 1976, Comrie 9 1989: 128, Corbett 2000: 56), asserting that in such languages, the number subsystem with the 10 most number values should always function at the top of the Animacy Hierarchy (2000: 92). A 11 further assumption of the Animacy Hierarchy is that splits between number subsystems should 12 fall between adjacent entries on the Animacy Hierarchy. This detailed case study of splits 13 between number subsystems in the Papuan language Nungon shows that the Animacy Hierarchy 14 fails to account for the distribution of number subsystems there in three ways. First, the 15 assumption that the subsystem with the most values must function at the top of the Animacy 16 Hierarchy falters. Second, Nungon number subsystems are split both within a single paradigm 17 and between multiple paradigms belonging to the same Animacy Hierarchy category. Third, 18 splits between these subsystems also occur in a category, tense marking, that is apparently 19 unrelated to the Animacy Hierarchy. Representation of number system splits as they relate or do 20 not relate to the Animacy Hierarchy requires a three-dimensional schema rather than the linear 21 Animacy Hierarchy. Finally, Nungon presents a previously-unattested dependency between 22 number and other grammatical systems. 23 24 Keywords: number, split, animacy, hierarchy, tense

25 1 Number systems and the Animacy Hierarchy

26 The grammars of some languages include two or more number subsystems that function in 27 different morphosyntactic contexts.1 These subsystems may involve different number value 28 distinctions: one subsystem could differentiate singular, dual, paucal, and , while another 29 might distinguish just singular and non-singular. Though not particularly rare, such systems have 30 received little attention by typologists. Dixon (2012: 52–55) calls them “mixed,” and presents 31 less a survey of these systems than an for appropriate descriptive terminology for 32 them. Dixon notes a tendency, not without exceptions, for the subsystem with the most number 33 values to apply to personal pronouns (2012: 55). Corbett (2000: 89–94, 120–124) discusses these 34 systems in the context of the nominal Animacy Hierarchy (Smith-Stark 1974, Silverstein 1976,

1 The of number is defined here following Dixon (2012: 473): “A grammatical system referring to quantity of referents.” Grammatical system, in turn, may be defined as a “closed set of choices one of which must be selected for a construction of a certain type” (Aikhenvald & Dixon 2011: 171). Following Corbett (2000), a number system is taken here to involve one or more number values, distinguished through formal marking of nominals and/or on modifiers and/or predicates.

35 Comrie 1989: 128, Corbett 2000: 56), making stronger predictions than Dixon as to which 36 number subsystems should function over which ranges of the hierarchy. Corbett (2000: 120–124) 37 also distinguishes between subsystems that are widely attested as sole number systems in other 38 languages (such as the widespread singular, dual, plural) and subsystems that are not attested as 39 the sole number system of other languages (such as the “conflated” subsystem that contrasts 40 ‘singular or dual’ and ‘plural’ for some inanimate nouns in the Otomanguean language Pame). 41 This paper introduces the term “split number” for number systems comprising 42 subsystems with different values. This term is appropriate because the number subsystems are 43 generally in complementary distribution, so that the researcher can pinpoint “splits” between the 44 subsystems. The term “split number” highlights the similarity between such complex number 45 systems and the splits in other languages’ grammars in ergativity marking (“split ergativity”; 46 Silverstein 1976), noun classes (“split agreement,” “split gender,” or “nominal” versus 47 “pronominal” noun classes; Heine 1982, Aikhenvald 2000: 67–80), and even constituent order 48 (“split word order”; Childs 2005). While Corbett (2000) and Dixon (2012) make some 49 generalizations about split number systems, the current study is the first to thoroughly examine 50 all morphosyntactic splits in the number system of a single language. This case study of the 51 Papuan language Nungon is thus able to correct and expand on earlier predictions. 52 Corbett’s typology of split number systems centers around what he and others (e.g. 53 Comrie 1989: 128) have called an Animacy Hierarchy. This particular hierarchy is a prominence 54 hierarchy (Lockwood & Macaulay 2012) ranking the personal pronouns with other nominals. 55 The form of the Animacy Hierarchy used by Corbett (2000) to build a typology of number 56 systems is based on that introduced by Smith-Stark (1974), which drew on cross-linguistic 57 evidence for which types of nouns are more likely to be marked for number. The Animacy 58 Hierarchy used for the typology of number differs slightly from the similar hierarchy related to 59 ergativity splits introduced by Silverstein (1976). Figure 1 is based on Corbett’s Animacy 60 Hierarchy (2000: 56): 61 pronouns nouns 1 > 2 > 3 > kin > human > animate > inanimate (speaker) (addressee) 62 Figure 1. The Animacy Hierarchy (Corbett 2000: 56) 63 64 One might argue that the term at the right end (‘bottom’) of the hierarchy could be further 65 divided into inanimate concrete nouns and inanimate abstract nouns, but data on distinctions in 66 number marking possibilities between these two may not be available in all reference grammars. 67 In his brief discussion of split number systems, Corbett assumes that the subsystem with 68 the most choices in number values must apply at least to the top of the hierarchy (Corbett 2000: 69 92). A tacit assumption about split systems inherent to the Animacy Hierarchy is that splits 70 between number systems should fall between adjacent entries on the Animacy Hierarchy. These

71 two conjectures will be referred to here as the Most Values at the Top and Split Along the 72 Hierarchy conjectures. 73 Through detailed analysis of the entire morphosyntactic system of one language, Nungon, 74 this paper provides a case study of the complex relations between multiple number subsystems in 75 a single grammar. in Nungon can be analyzed as containing three number 76 subsystems: a maximal subsystem with three number values distinguished (singular, dual, and 77 plural), a medial subsystem with two number values distinguished (singular, non-singular), and a 78 minimal subsystem in which there is no number value distinguished.2 In Nungon, splits between 79 different number subsystems occur not only in expected locations along the Animacy Hierarchy, 80 but also between different sets of personal pronouns, between different verbal person/number 81 affixes, and between different verbal tense suffixes. There are multiple motivations for these 82 splits which only partially subsume the Animacy Hierarchy. After section 2, which gives 83 background on the Nungon language, section 3 outlines the number split between sets of 84 personal pronouns. Section 4 presents the number splits among nouns. The number splits 85 between Nungon pronouns and nouns are then compared with predictions of the Animacy 86 Hierarchy in section 5; modifications of the Animacy Hierarchy are suggested. Section 6 87 describes and analyzes the highly unusual number split between verbal tense suffixes. Section 7 88 evaluates diachronic explanations for the number splits between personal pronoun sets and tense 89 suffixes. Section 8 concludes the paper by reconsidering all the number splits in Nungon 90 morphosyntax together, and comparing dependencies between Nungon grammatical systems to 91 those predicted by Aikhenvald & Dixon (1998, 2011).

92 2 The Nungon language: background and typological profile

93 Claassen & McElhanon (1970) classed the Papuan Uruwa language family, which includes 94 Nungon, as Finisterre within the larger Finisterre-Huon language group (Madang and Morobe 95 Provinces, Papua New Guinea). Nungon is spoken by about 1,000 people; it is the main language 96 spoken in homes and the first language of all children born to Nungon-speaking parents. 97 National lingua franca Tok Pisin is used in the government school (grades 3–8) and in some 98 church contexts. Nungon is agglutinating, with some fusion. Constituent order is verb-final. The 99 language is predominantly suffixing, but verbal objects are requisitely marked on some verbs as 100 prefixes. Like many other Papuan languages (Roberts 1997), Nungon features clause chaining, 101 with switch-reference marked on medial verbs. The first grammatical description of Nungon is 102 Author (2014c, 2017a). Nungon examples throughout the paper are drawn from a 146,700-word 103 corpus of transcribed Nungon narratives and dialogues. While there are four major dialects of

2 This paper follows Corbett (2000: 120-121, 173) in considering the subsystem in which number is not marked as itself a number system, rather than representing the inapplicability of number to some area of grammar. This can be compared with the analysis of zero marking in a morphological paradigm: the null number subsystem is contrastive (in complementary distribution) with the other subsystems. As Corbett puts it, such a null subsystem typically applies to nouns “which are off the end of the scale for number-differentiability” (2000: 173).

104 Nungon (itself the southern half of an oval dialect continuum—see Author 2014b); the Towet 105 village dialect is the basis for all discussion here, unless otherwise noted.3 106 Number in Nungon is marked in various ways on pronouns, nouns, and verbs. When 107 reduplicated, some adjectives convey both non-singularity and intensified meaning (Author 108 2017a, b). As noted in section 1, Nungon has a split number system, with three different number 109 subsystems functioning in different areas of the Nungon grammar. These subsystems are: a 110 tripartite subsystem, containing the number values singular, dual, and plural (where plural 111 indicates more than two referents); a bipartite subsystem, containing the number values singular 112 and non-singular (where non-singular indicates more than one referent); and a null system, in 113 which number is not marked. Crucially, these number subsystems are in complementary 114 distribution: where one applies, the other two do not.

115 3 Number system splits among pronouns

116 The top three categories of Corbett’s version of the Animacy Hierarchy in Figure 1 are personal 117 pronouns. Nungon grammar includes two sets of free personal pronouns: a neutral set, and a set 118 used for and reflexivity. In the Finisterre-Huon Papuan linguistic tradition (McElhanon 119 1973, inter alia), the neutral pronouns are called “basic” and the contrastive/reflexive pronouns 120 are called “emphatic.” Both of these can further combine with enclitics marking grammatical 121 relations and topicality. Nungon can also be analyzed as having bound personal pronouns (Dixon 122 2010: 116); on verbs, -indexing suffixes and -indexing prefixes can be called bound 123 pronouns, as can the pertensive (possessive) markers on nouns, which follow the Possessed noun 124 and index the person and number of the Possessor. All three Nungon number subsystems— 125 tripartite, bipartite, and null—are represented among the free personal pronouns. Among the 126 bound personal pronouns, the maximal, tripartite subsystem and the medial, bipartite subsystem 127 are both represented.

128 3.1 Free pronouns

129 The two sets of Nungon free personal pronouns are shown in Table 1 and Table 2. 130 singular non-singular 1 nok non 2 gok hon 3 yu 131 Table 1. Nungon basic pronouns 132 133 134 135

3 Dialects mostly vary in lexicon and phonology. The Towet village variety has 14 phonemic consonants and 6 phonemic vowels, with contrastive vowel length.

singular dual plural 1 naga nori noni 2 gaga hori honi 3 ino yori yoni 136 Table 2. Nungon emphatic pronouns 137 138 As seen in Table 1, two number subsystems are evident among the Nungon basic 139 pronouns: the first and second person forms encode number according to the bipartite subsystem, 140 while the third person form operates according to the null system, with no number values 141 distinguished. In contrast, Table 2 shows that all Nungon emphatic pronouns encode number 142 according to the tripartite subsystem. Section 7 introduces comparative data that suggests that 143 historical developments led to a reduction in number value distinctions in the first and second 144 person basic pronouns. The rest of this section presents the functions and distribution of Nungon 145 free pronouns; these details may also help explain the number splits between the two sets and 146 within the basic pronouns. 147 Nungon transitive and intransitive subject arguments, including personal pronouns, are 148 omissible in clauses with verbal predicates; the inflected verb obligatorily indexes the person and 149 number of its subject argument. See §3.2 for more on object argument person/number indexation 150 with some verbs. Free personal pronouns typically refer to humans and other animates, while 151 inanimates may be referred to with nominal demonstratives. Free pronouns cannot be modified 152 by descriptive adjectives, although they may be modified by cardinal numerals. 153 When they occur and their referents are already indexed on verbs, the two sets of free 154 pronouns serve different functions. The basic set add emphasis or mild contrast, while the 155 emphatic set function reflexively or autoreflexively, or contribute strong contrast. This is shown 156 in examples (1) and (3).

157 1) Gok ma=ng-i-ya, non ma=ngo-ni-n. 158 2SG.PRO NEG=go-DS.2SG-MV 1NSG.PRO NEG=go-IRR.PL-1NSG 159 If you don’t go, we won’t go. [Literally: ‘You not having gone, we won’t go.’]

160 The basic pronouns could be omitted in (1) with the subject argument referents understood, but 161 some of the contrast would be lost. 162 The basic pronouns are the functionally-unmarked shifters employed in verbless clauses, 163 as in (2):

164 2) Gok, oe. 165 2SG.PRO woman 166 You are a woman. [Literally: ‘you, woman’]

167 The emphatic pronouns can have contrastive effect, as in (3), reflexive effect, as in (4), or 168 autoreflexive effect, as in (5). Example (3) includes a coreferential basic pronoun and emphatic 169 pronoun:

170 3) Gebi=rot, non urop, nori=ho=gon 171 Gaby=COMIT 1NSG.PRO enough 1DU.EMPH.PRO=FOC=RSTR 172 173 yonggut-do-mok. 174 laugh-RP-1DU 175 Gaby (and I) then laughed, just the two of us. [Literally: ‘Along with Gaby, we then 176 laughed, just we two.’]

177 Here, the basic pronoun contributes little to no contrast, while the emphatic pronoun is used for 178 strong contrast between the referents of the pronoun and other people present at the time of 179 reference. Note that because two different number subsystems apply to the first person basic and 180 emphatic pronouns, the potential scope of the basic pronoun (non-singular) is greater than that of 181 the emphatic pronoun (dual). Example (4) shows reflexive use of an emphatic pronoun:

182 4) Nok naga n-et-ta-t. 183 1SG.PRO 1SG.PRO.EMPH 1SG.O-beat-PRES.SG-1SG 184 I hit myself.

185 Autoreflexive function is indicated with the addition of one of the three autoreflexive 186 clitics, =wut ‘by one’s own power,’ =nang ‘unaccompanied (by oneself),’ and =gon ‘just 187 (oneself),’ as shown in (5):

188 5) Naga=wut hagam towi-wa-t. 189 1SG.PRO.EMPH=AUTOREFL stairs fix-NP.SG-1SG 190 I fixed the stairs myself.

191 While free pronouns are omissible from a clause when they are coreferential with 192 subjects indexed on the verb, they are obligatory when they are the sole indicators of an object or 193 oblique argument in a clause. Pronouns as objects most often occur preceding the verb, as in (6):

194 6) Nok honggit-do-k. 195 1SG.PRO grab-RP-3SG 196 He caught me.

197 The most common verbs that often take people as object arguments, such as yandi- ‘show,’ ino- 198 ‘tell,’ imo- ‘give,’ aa- ‘see,’ wet- ‘beat,’ and light verb constructions such as yeng ho- ‘pass’ and 199 wep mo- ‘touch,’ all include indexation of the object person and number in a bound pronominal 200 prefix (see next section). This makes free pronouns indicating the object person and number 201 optional with these predicates.

202 3.2 Bound pronouns

203 Nungon has three types of bound pronouns (dependent shifter morphemes): object argument- 204 indexing verbal prefixes, subject argument-indexing verbal suffixes, and pertensive markers on 205 nouns, which follow the Possessed noun and index the person and number of the Possessor. The 206 object-indexing prefixes operate according to the bipartite number subsystem, while the bound

207 pronouns that reference subject arguments and Possessors operate according to the tripartite 208 subsystem. This split among bound pronouns could relate to historical constituent order; both 209 subject indexers and pertensive markers are suffixed, while object indexers are prefixed. 210 The object prefixes that occur on a closed class of 13 transitive verbs and on preverbal 211 elements in two light verb constructions are shown in Table 3. The vowel of these prefixes varies 212 depending on the verb, but the initial consonant is stable for all object person/number 213 combinations except 3sg, which shows some variation from verb to verb. 214 singular non-singular 1 n(a)- n(i)- 2 g(a)- k(a)- 3 i-/y-/w-/Ø/t- y(o)- 215 Table 3. Object person/number prefixes 216 217 In Nungon, only a closed class of 13 transitive verbs and the preverbal elements in two 218 light verb constructions feature obligatory prefixes indexing the person, or person and number, 219 of the object argument (Author 2014a, 2014c, 2017a). All other verbs never take such prefixes. 220 Many of the verbs that bear prefixes indexing both person and number of the object argument 221 may be understood to be verbs with prototypically human, or at least prototypically animate, 222 object arguments. An exhaustive list of verbs that take object prefixes is: aa- ‘see,’ imo- ‘give,’ 223 ino- ‘tell,’ taambit- ‘tread on,’ tan- ‘follow,’ temo- ‘shoot,’ wet- ‘beat,’ yama- ‘look after,’ yandi- 224 ‘show to,’ yangat- ‘escort,’ yii- ‘bite,’ yö- ‘set (s.o.) down,’ and yuu- ‘roll, lead on.’ The two 225 light verb constructions that also include object indexation prefixed to the preverbal element 226 likely originated as serial-like tight multi-verb constructions, with the preverbal element a verb: 227 wep mo- ‘touch’ and yeng hot- ‘pass.’ 228 The verb yö- ‘set (s.o.) down,’ for instance, is used only with animate object arguments. 229 A lexically-equivalent verb hi- ‘put down,’ which never bears object prefixes, is used with 230 inanimates. Placing a child on the ground is described with yö-, as is letting vehicle passengers 231 off at a certain point. This is seen in example (7), which refers to a child of about six years:

232 7) Nok Gorungon honggir-a, to-ng yöng-ka-t. 233 1SG.PRO Gorungon grab-MV SG.O.take-DEP 3SG.O.put.down-PRES.SG-1SG 234 Grabbing Gorungon, I take him and place him down.

235 In contrast, placing a bag of vegetables on the ground is described with hi-. Dogs and pigs—both 236 culturally-important, domesticated animals—are placed using yö-. When asked, speakers agreed 237 that a small child would be placed using yö- unless it were an infant incapable of moving itself; 238 putting down such an infant could also be described with hi-. Insects such as butterflies would 239 probably be placed using hi-, and human or mammalian corpses would be placed with hi-. Other 240 verbs with obligatory object person/number prefixes cannot be argued as easily to typically take 241 animate objects: aa- ‘see, look at,’ yuu- ‘roll (from side to side),’ taambit- ‘tread on,’ wet- ‘beat.’

242 For the closed class of 13 verbs, these prefixes are obligatory, even when an explicit 243 object argument is also present in the clause. Example (4) showed the verb wet- ‘beat’ bearing 244 the 1sg object prefix, with the 1sg emphatic pronoun also present as explicit object argument. 245 Another example, where the basic 1sg pronoun functions as explicit object argument and 246 co-occurs with an object prefix on the verb, is in (8).

247 8) Böiwa=ho nok na-no-go-k. 248 Böiwa=FOC 1SG.PRO 1SG.O-tell-RP-3SG 249 Böiwa told (it) to me.

250 It is likely that the object person/number prefixes originated as free personal pronouns, 251 but this may have occurred in proto-Finisterre-Huon. Nungon is a verb-final language, and (as 252 noted in section 3.1) the preferred location of an explicit object argument is immediately before 253 the verb. of Table 3 with Tables 1 and 2 shows that the object prefixes, with the 254 exceptions of 3sg and 2nsg, begin with the same segments as the corresponding basic or 255 emphatic personal pronouns. If the free pronouns at the time of such cliticization had the forms 256 they do in Nungon today, however, the vowel distinction between 1sg and 1nsg object prefixes 257 would be unmotivated. Section 7 gives evidence that the object prefixes date to proto-Finisterre- 258 Huon. 259 Unlike the verbal object prefixes, verbal subject person/number suffixes operate 260 according to a maximal, tripartite number subsystem. There are two main formal groupings of 261 these suffixes (Author 2015), depending on whether the suffix follows a tense suffix or not; the 262 tripartite number subsystem applies to both groupings. The suffixes employed after tense 263 markers are in Table 4, while those that directly follow the verb root in the absence of any tense 264 marker are in Table 5. The suffixes in Table 4 occur with tense- or reality-status-marked final 265 verbs—the fully inflected verbal predicates of independent verbal clauses or the last clause of a 266 clause chain. The suffixes in Table 5 occur on verbs in that lack a tense suffix after 267 the verb root; these are the Immediate Imperative , the Counterfactual inflection, and 268 Different-Subject medial verbs in non-final clauses in a clause chain. 269 singular dual plural 1 -t -mok -mong 2 -rok -morok -ng 3 -k 270 Table 4. Subject suffixes (post-tense suffix) 271 singular dual plural 1 -wa/-e -ra -na 2 -i -un -u 3 -un 272 Table 5. Subject suffixes (post-verb root)

273 In both Tables 4 and 5, there are single forms for second and third person dual and plural, 274 creating a formal distinction in those numbers between first person and non-first person. 275 Syncretism between second and third persons in dual and plural numbers is characteristic of 276 many Papuan languages, especially Papuan languages of the Finisterre-Huon group (Foley 1986, 277 McElhanon 1973). 278 Since the verbal subject and object affixes operate according to different number 279 subsystems, verbs including bound pronouns with different number subsystems are 280 commonplace. Example (9) shows this with two medial clauses from a portion of a clause chain 281 describing traditional trading practices between Saruwaged Mountains communities like Nungon 282 speakers and the coastal Vitiaz Strait communities:

283 9) Yu=ho uwa yo-m-u-ya, yu=ho tik nungon 284 3.PRO=FOC pot 3.NSG.O-give-DS.2/3PL-MV 3.PRO=FOC bark.cloth what 285 286 yo-m-u-ya… 287 3.NSG.O-give-DS.2/3PL-MV 288 Theyi having given themj pots, theyj having given themi bark-cloth and whatnot…

289 Here, the coastal people give the mountain people clay pots, and the mountain people give them 290 bark-cloth and other handicrafts from their region. In the first clause, the party of coastal people 291 is referenced by the third person basic pronoun yu, which operates according to the null number 292 subsystem, and by the 2/3pl subject suffix -u on the verb ‘give,’ which operates according to the 293 tripartite number subsystem. In this clause, the mountain people are referenced by the 3nsg 294 object prefix yo- on the verb ‘give,’ which functions according to the bipartite number 295 subsystem. This means that at this point, the morphosyntax allows for the possibility that the 296 mountain people could be a party of only two. In the second clause, the roles reverse, and it is the 297 mountain people who are referenced by the basic third person pronoun yu and the 2/3pl subject 298 suffix on ‘give’; now it is clear that they form a party of at least three, as was established for the 299 coastal people in the preceding clause. Note also that the inanimate nouns uwa ‘pot’ and tik 300 ‘bark-cloth’ are unmarked for number (see section 4 on number in nouns). 301 The pertensive markers, which index the person and number of the Possessor in 302 possessive NPs, take two forms. The singular pertensive markers are suffixes that attach to the 303 Possessed noun, while the dual and plural pertensive markers are phonologically free words that 304 immediately follow the Possessed noun. The pertensive markers used with dual and plural 305 Possessors further differ in three ways from the markers used with singular Possessors: a) dual 306 and plural pertensive markers are formally identical to dual and plural emphatic free personal 307 pronouns, while singular pertensive markers are not transparently related to free personal 308 pronouns; b) dual and plural pertensive markers are disyllabic, while singular pertensive markers 309 are monosyllables; c) singular pertensive markers further host number markers indexing the 310 number of the Possessed (see section 4), while this is impossible with dual and plural pertensive 311 markers. 312

singular dual plural 1 -na nori noni 2 -a hori honi 3 -(n)o yori yoni 313 Table 6. Pertensive markers 314 315 In Nungon grammar, the only formal distinction between alienable and inalienable 316 arises when the Possessor is 3sg, and the noun referring to the Possessed has a 317 particular phonological form. If a noun is phonologically eligible for formal marking of 318 alienability, the distinction is between 3sg pertensive marker -o for inalienable and -no for 319 alienable possession. Each noun is inherently either alienably or inalienably possessable. Such a 320 formal distinction is not made with any of the other pertensive markers. 321 Example (10) shows the 1du pertensive marker with the kin term owi ‘grandmother’:

322 10) Owi nori=ho ni-ma-ng-a it-do-k. 323 grandmother 1DU.POSS=FOC 1NSG.O-look.after-DEP-MV be-RP-3SG 324 Our grandmother used to look after us.

325 Here, the pertensive marker indexes number according to the tripartite subsystem, while the 326 coreferential object prefix on the verb ‘look after’ indexes number according to the bipartite 327 system. 328 329 Number system splits among Nungon pronouns are summarized in Table 7. 330 pronoun set number subsystem emphatic free pronouns, pertensive markers, TRIPARTITE subject-indexing verbal suffixes 1st and 2nd person basic free pronouns BIPARTITE object-indexing verbal prefixes

3rd person basic free pronoun NULL 331 Table 7. Number system splits among Nungon free and bound pronouns 332 333 While transitive verbs that cannot take object prefixes at all could be said to lack number 334 marking for objects, thus featuring a null number subsystem for object indexation, they also lack 335 any person marking for objects. Thus, they are not precisely equivalent to zero-marked terms 336 within the same prefixing paradigm. For this reason, these verbs are excluded from discussion 337 here in terms of a number subsystem. 338 339

340 4 The number system split among nouns

341 Overt number marking on nouns encompasses purely grammatical number marking and several 342 more lexically-entwined indicators of number: three different associative markers, a special 343 marker for non-singular cohorts of same-generation blood relatives, and repetition of certain 344 modifying adjectives for both intensification and non-singularity. All number marking on nouns 345 accords with the predictions of the Animacy Hierarchy. Grammatical number marking is 346 discussed first, followed by the other types. 347 A closed sub-class of 37 Nungon nouns may be marked for grammatical number; this 348 includes all 32 consanguineal and affinal kin terms, plus five additional nouns that prototypically 349 have human referents (nuk ‘friend, comrade’ morum ‘owner,’ wase ‘namesake,’ oruk oruk 350 ‘friend, classmate,’ and nip oruk ‘trade-friend’). These nouns are only marked for number, 351 according to the tripartite subsystem, when they bear a singular pertensive suffix (see Table 6). 352 All other nouns—and nouns of this sub-group if they lack a singular pertensive suffix—are never 353 marked for number. Number marking on the noun nuk ‘friend, comrade’ is shown in Table 8. 354 Possessor singular Possessed dual Possessed plural Possessed 1sg nuk-na ‘my friend’ nuk-na-in ‘my (2) friends’ nuk-na-i ‘my friends’ ‘your ‘your (2) 2sg nug-a nug-a-in nug-a-i ‘your friends’ friend’ friends’ ‘his/her ‘his/her (2) 3sg nuk-no nuk-n-in nuk-n-i ‘his/her friends’ friend’ friends’ 355 Table 8. Prototypically-human noun nuk ‘friend, comrade’ with number marking4 356 357 All kin terms behave like nuk in Table 8 in being obligatorily marked for number when 358 bearing a singular pertensive suffix. A minor distinction arises in the form of the 3sg pertensive 359 suffix; nuk ‘friend’ is always alienably possessed, and the forms in Table 8 with 3sg Possessor 360 include -no or the reduced -n indexing 3sg. For an inalienably possessed noun like the kin term 361 oruk ‘brother of male,’ the forms lack this /n/, as seen in Table 9. 362 363 364 365 366 367 368

4 Note that another morpheme break could be written within the dual number suffix here, as -i-n, showing the formal relationship between the plural suffix -i and the dual suffix -i(-)n. It is not uncommon for a dual marker to be morphologically based on a plural marking; this is found in Chumash (Kroeber 1904:33, cited in Croft 2003:127), and Manam (Lichtenberk 1983:267, cited in Croft 2003:126).

Possessor singular Possessed dual Possessed plural Possessed ‘my ‘my (2) oruk-na- 1sg oruk-na oruk-na-in ‘my brothers’ brother’ brothers’ i ‘your ‘your (2) 2sg orug-a orug-a-in orug-a-i ‘your brothers’ brother’ brothers’ ‘his ‘his (2) 3sg orug-o orug-in orug-i ‘his brothers’ brother’ brothers’ 369 Table 9. Kin term oruk ‘brother of male’ with number marking 370 371 Examples (11) and (12) show the kin term bap ‘maternal uncle’ with and without dual 372 number marking:

373 11) Bap-na-in na-a-morok. 374 maternal.uncle-1SG.POSS-DU eat-NP.NSG-2/3DU 375 My two maternal uncles ate. 376 12) Bap noni na-a-morok. 377 maternal.uncle 1PL.POSS eat-NP.NSG-2/3DU 378 Our two maternal uncles ate.

379 In (11), bap ‘maternal uncle’ bears the singular pertensive suffix -na, which enables it to also be 380 marked for number. In (12), bap bears the plural pertensive marker noni, which does not allow 381 for further marking of number; the fact that there are two uncles is formally indicated only in the 382 verbal subject suffix -morok. 383 In contrast to (11), (13) shows the noun haa ‘area, environs’ with dual reference but 384 unmarked for number:

385 13) Haa yoi ir-a-morok: Yomöng, Hok Bun. 386 area two be-PRES.NSG-2/3DU Yomöng Hok Bun 387 Two places exist: Yomöng and Hok Bun.

388 Even nouns referring to higher animates with strong cultural significance and ties to humans, 389 such as hunting dogs, are not marked for number, as seen in (14):

390 14) Hap-na koit-no ir-a-ng. 391 dog-1SG.POSS many-ADJ be-PRES.NSG-2.3PL 392 My dogs are many.

393 Here, hap-na bears the 1sg pertensive suffix -na, like nuk-na-i ‘my friends’ in Table 8, but 394 because hap is neither a kin term nor a noun with prototypically-human reference, it does not 395 further receive the plural number suffix -i. The noun here is marked according to the null number 396 subsystem, while its referent is also indexed in the verbal subject suffix according to the tripartite 397 number subsystem.

398 While most of the Nungon kin terms and the other five prototypically-human nouns 399 usually refer to humans, some can be extended to non-humans, even inanimates. When 400 occasionally used to refer to non-humans, these 37 nouns are still marked for number. For 401 instance, in the Nungon taxonomical system, a genetic relationship between species of flora or 402 fauna is described using the term nuk ‘friend’. In example (15), nuk refers to trees, not humans, 403 but is still marked for plural number:

404 15) Giyöng=kon nuk-n-i koit-no ir-a-ng. 405 betelnut=GEN friend-3SG.POSS-PL many-ADJ exist-PRES.NSG-2/3PL 406 Many relatives of betelnut exist.

407 The split between number subsystems applicable to Nungon nouns based on purely 408 grammatical marking of number is summarized in Table 10. 409 nominal sub-class number subsystem

prototypically-human, bearing singular pertensive suffix TRIPARTITE (SG./DU./PL.)

all others NULL 410 Table 10. Number subsystems with Nungon nouns 411 412 The rest of this section examines five additional ways to indicate non-singularity on 413 Nungon nominals that could be seen as operating outside the domain of grammatical number. A 414 full description of quantification in Nungon is in Author (2017b). 415 Nungon has three markers of associative non-singular (Moravcsik 2003). Associative 416 constructions in Nungon combine a personal name or epithet for a specific person with a marker 417 to indicate a group formed of that person and one or more associates. The Nungon associative 418 markers are termed: the pertensive associative, the couple associative, and the general 419 associative. These are presented in Table 11. 420 associative type form meaning number reference kin term with 3sg pertensive suffix ‘X along with pertensive fused to Comitative marker his/her [kin two or more term]s’ noun oemna ‘couple’ follows the ‘X and spouse’ couple two name of one spouse word gomong follows the name(s) ‘X (and Z) and general two or more of one or more of the associates associates’ 421 Table 11. Nungon associative plural markers 422 423

424 The general associative marker is illustrated in example (16):

425 16) Lisa Sekari gomong yoo-ng-a ep-bo-k. 426 Lisa Zachary ASSOC NSG.take-DEP-MV come-RP-3SG 427 (The airplane) came bringing Lisa, Zachary and their group.

428 The fourth lexically-entwined way in which nouns can be marked for number is a marker 429 of non-singularity used only with reciprocal, same-generation, consanguineal kin terms (Author 430 2014c, 2017a: 177-182, 451-452). The eligible kin terms are: naat ‘different-sex sibling,’ daa 431 ‘sister of female,’ oruk ‘brother of male,’ and nip ‘cross-cousin.’ When one of these is preceded 432 by the word yomot, the resulting noun phrase indicates a group of two or more mixed-sex 433 siblings, sisters, brothers, or cousins. This is shown in example (17):

434 17) Mak nori yomot daa. 435 mother 1DU.POSS COLL sister.of.female 436 Our mothers are sisters.

437 A fifth way in which non-singular number can be covertly indicated for a noun phrase is 438 through repetition of an eligible modifying adjective (Author 2017a: 215-218). This process 439 obligatorily also contributes intensification of the adjective’s semantics, however, so it is not 440 solely related to number. Adjectives observed repeated for intensification and non-singularity all 441 come from the size and value semantic groups; these adjectives are: opmou ‘small,’ morö ‘large,’ 442 orog-o ‘good, beautiful,’ moin-no ‘bad, ugly,’ and imbange ‘wonderful, beautiful.’ These have 443 been observed modifying nouns with human reference, such as gungak ‘child,’ as well as 444 inanimate nouns such as böörong ‘rock’ in (18):

445 18) böörong opmou opmou 446 rock small small 447 very small rocks

448 As described in Author (2017a: 216), the noun phrase in (18) can neither refer to a single small 449 rock or to several medium-sized rocks; both non-singularity and intensification are obligatory 450 readings. Such a noun phrase can further bear a numeral modifier: ‘three very small rocks.’ 451 This last marginal way of indicating non-singularity is the only one considered in this 452 section that has a range across the Animacy Hierarchy; adjectival modifiers of nouns referring to 453 humans and stones alike can be repeated for non-singularity and intensification. It is as yet 454 unproven whether adjective repetition is acceptable with nouns referring to abstract concepts, 455 such as meep-mo ‘problem’ or turong-o ‘sin.’ 456 Corbett (2000: 101–111) considers associative constructions to be outside the sphere of 457 grammatical number. In Nungon, these constructions, along with yomot and adjective repetition, 458 are restricted to certain lexical items: all three associative markers require the use of a personal 459 name or epithet, with the pertensive associative further requiring a kin term; the marker yomot 460 can only occur with same-generation, self-reciprocal kin terms; and only certain adjectives can

461 be repeated when modifying a noun. The associative markers also inherently convey both 462 number (either non-singularity or duality) and semantics, since they specify the type of 463 relationship between members of a group. 464 Although these markings will not be explored further in the following discussion, it may 465 be noted that their application accords with the Animacy Hierarchy. The pertensive and couple 466 associative markers, and yomot, are restricted to kin, while the general associative can occur with 467 either kin or general human nouns. Indication of number through modifying adjective repetition 468 is an option for nouns across the entire range of the Animacy Hierarchy, from humans to 469 inanimates.

470 5 Nungon nominals and the Animacy Hierarchy

471 Plotting the number subsystems presented in Tables 7 and 10 for Nungon pronouns and nouns 472 along the Animacy Hierarchy in Figure 1 yields mixed results. If just emphatic pronouns are 473 considered along with nouns, this does not violate the two Animacy Hierarchy stipulations laid 474 out in section 1: the Most Values at the Top and Split Along the Hierarchy stipulations are 475 upheld, as seen in Figure 2. 476 pronouns nouns 1 > 2 > 3 > kin > human > animate > inanimate TRIPARTITE NULL 477 Figure 2. Nungon emphatic pronoun and noun number subsystems with the Animacy Hierarchy 478 479 The emphatic personal pronouns, as well as kin terms and non-kin human nouns, operate 480 with the maximal, tripartite number value subsystem. All other nouns are grammatically marked 481 for number according to the null subsystem. In Corbett’s words, they are “off the scale of 482 number differentiability” (2000: 121). Here, the subsystem with most number values applies to 483 the top of the hierarchy, and there is a monotonic decrease in number values per subsystem 484 across the hierarchy. 485 But when nouns are considered together with the basic independent pronouns, both 486 Animacy Hierarchy stipulations are violated, as seen in Figure 3. 487 pronouns nouns 1 > 2 > 3 > kin > human > animate > inanimate BIPARTITE NULL TRIPARTITE NULL 488 Figure 3. Nungon basic pronoun and noun number subsystems with the Animacy Hierarchy 489 490 In Figure 3, the Most Values at the Top stipulation is violated, since the range of the 491 number subsystem with the most number values falls in the middle of the hierarchy. The Most 492 Values at the Top stipulation is also violated if the bound object pronouns are plotted along with 493 nouns. When taken together, all categories considered thus far also violate the Splits Along the

494 Hierarchy conjecture. Figure 4 shows all number marking discussed thus far plotted along the 495 Animacy Hierarchy. 496 pronouns nouns kin human animate 1 > 2 > 3 > inanimate > > > basic BIPARTITE NULL — — pronouns emphatic TRIPARTITE TRIPARTITE — — pronouns bound object BIPARTITE BIPARTITE — — pronouns bound subject pronouns, TRIPARTITE TRIPARTITE — — pertensive markers possessed nouns — — TRIPARTITE NULL (grammatical marking) nouns TRIPARTITE OR (associatives — — NULL BIPARTITE and yomot) nominals (adjective — — BIPARTITE BIPARTITE repetition) 497 Figure 4. All Nungon pronoun and noun number subsystems with the Animacy Hierarchy 498 499 Figure 4 shows that splits between number subsystems among Nungon pronouns and 500 nominals do not all fall between categories on the Animacy Hierarchy. Within each paradigm— 501 among the basic pronouns, or among possessed nouns, for instance—the assumptions of the 502 Animacy Hierarchy are upheld. It is when these paradigms are combined that the Animacy 503 Hierarchy is both functionally and representationally inadequate. Full representation of the 504 Nungon split number system requires that an additional dimension be added to the linear 505 Animacy Hierarchy, so that number splits apparently unrelated to the Animacy Hierarchy can 506 also be charted. An initial attempt at this is in Figure 5. 507 508 509

number splits within paradigms 1 > 2 > 3 > kin > human > animate > inanimate basic

pronouns

number splits emphatic

across paradigms pronouns object

prefixes

subject

suffixes

nouns 510 Figure 5. Planar schema showing Nungon number splits on two axes 511 512 Figure 5 adds another axis to the linear Animacy Hierarchy so that both number splits within and 513 between pronoun paradigms can be represented. Here, splits between number subsystems are 514 indicated with vertical and horizontal dotted lines. Shaded areas are not applicable to a given 515 category. The next section will add a further category—and another axis.

516 6 The number splits among tenses

517 Most unusual about Nungon and other Finisterre Papuan languages is the distribution of number 518 subsystems in double-marking of subject number in verbs marked for tense. In Nungon, subject 519 number is indexed both in the final person/number suffix (Table 4) and, in all tenses except the 520 Remote Past, again in the tense suffix, where it is fused with tense marking. Different number 521 subsystems are evident in different tense markers. The splits between these number subsystems 522 coincide with the division of tenses under negation, when two tense distinctions are neutralized. 523 More on dependencies between grammatical systems in Nungon (negation and tense, tense and 524 number, number and person) is in section 9. This example of number system splits relates to 525 what Corbett terms “target number,” rather than “controller number,” since it indexes the number 526 of an explicit or non-explicit subject argument (2000: 178). The number splits on tense suffixes 527 are independent of controller (subject argument) number. 528 As noted in section 3.2, the Nungon Immediate Imperative, Counterfactual, and Medial 529 verb forms are unmarked for tense. Verbal inflections where tense is marked are: Remote Past, 530 Near Past, Present, Near Future, Remote Future, and two inflections that are formally related to 531 the Remote Future tense but relate to reality status or mood: the Future Irrealis and Delayed 532 Imperative. Time frames in which the five Nungon tenses are applied in positive polarity are 533 shown in Figure 6. 534

Before Between now Tomorrow and yesterday Yesterday Earlier today Right now and later today beyond PRESENT REMOTE PAST (some PRESENT NEAR FUTURE REMOTE FUTURE verbs) NEAR PAST 535 Figure 6. use relating to relative time, positive polarity 536 537 As seen in Figure 6, both Near Past and Remote Past can be used to describe events 538 ‘yesterday,’ while only the Remote Past can be used for events prior to ‘yesterday,’ and only the 539 Near Past for events ‘earlier today.’ The cutoff point between use of the Near Future and the 540 Remote Future is the end of ‘today.’ Of course, where ‘today’ ends is not reckoned in terms of 541 absolute time, sunrise or sunset, in the Nungon tense system: those concepts reside in a more 542 lexically-determined domain. Rather, the boundary between tenses largely relates to whether a 543 speaker anticipates sleeping for a regular night’s sleep before the events described. A Nungon 544 speaker about to bed down for the night speaks about events anticipated for the next morning 545 using the Remote Future. But a speaker who wakes in the middle of the night might refer to the 546 morning’s events either in the Near Future or Remote Future (Author 2014c, 2017). 547 In verbal inflections with a tense suffix, morpheme ordering is as in Figure 7. Elements in 548 parentheses are not present in all seven inflections. 549 (object tense / subject (negating (Remote person/number verb root (number) person/number proclitic) Future suffix)5 prefix) suffix suffix 550 Figure 7. Verb morpheme ordering when tense marking is present 551 552 Figure 7 includes three morphological slots in which number is marked. As described in 553 section 3.2, the object prefix indexes the number (and person) of the verb’s object argument for a 554 closed class of transitive verbs. This prefix encodes number according to the bipartite subsystem, 555 while the subject person/number suffix encodes subject number according to the tripartite 556 system. The third slot in which argument number is indexed is the tense/number suffix, which 557 indexes subject number; here, number is fused with tense, rather than with person. The number 558 subsystem operating here depends on the tense, with all three number subsystems used. 559 Tense suffixes vary slightly in form based on the morphological class of the verb root 560 they follow (details in Author 2014a, 2014c, 2017); the forms in Table 12 are the forms used 561 with verb roots of the H-class, the morphological verb class with the most members.

5 The tense/number suffixes in the slot after the verb root are identical for the Remote Future, Future Irrealis, and Delayed Imperative. The Remote Future is distinguished from these inflections in positive polarity by the unchanging Remote Future suffix -ma, which follows the subject person/number suffix.

562 tense suffix form number subsystem Remote Past -go- (all numbers) NULL -wa- (singular) Near Past -a- (non-singular) BIPARTITE -ha- (singular) Present -wa- (non-singular) -wangka- (singular) Near Future -rangka- (dual) -nangka- (plural) TRIPARTITE -i- (singular) Remote Future -ri- (dual) -ni- (plural) 563 Table 12. Tense suffix forms after H-class verbs 564 565 The full paradigm showing all tense inflections with all person/number subject suffixes 566 for the H-class verb na- ‘eat’ is in Table 13. 567 1sg 2sg 3sg 1du 2/3du 1pl 2/3pl na-go- na-go- na-go- RP na-go-t na-go-rok na-go-k na-gu-ng mok morok mong na-a- na-a- NP na-wa-t na-wa-rok na-wa-k na-a-mok na-a-ng morok mong na-wa- na-wa- na-wa- PRES na-ha-t na-ha-rok na-ha-k na-wa-ng mok morok mong na- na- na- na- na-wangka- na-wangka- na- NF rangka- rangka- nangka- nangka- t rok wangka-t mok morok mong ng na-ri- na-i-rok- na-ri-n- na-ni-n- na-ni-ng- RF na-i-t-ma na-i-k-ma morok- ma ma ma ma ma 568 Table 13. Full paradigm for na- ‘eat’ 569 570 As seen in Tables 12 and 13, the Remote Past suffix does not change forms depending on 571 the verbal subject argument number; the Near Past and Present have two forms, one for singular 572 and the other for non-singular subject arguments; and the Near Future and Remote Future have 573 three forms, one for singular, one for dual, and one for plural subject arguments. The splits 574 between number subsystems here align with patterns of neutralization under negation. This is 575 articulated in Table 14. 576

tense number subsystem neutralizing behavior under negation Remote Past NULL none Near Past BIPARTITE Near Past form used Present Near Future Future Irrealis form used (with TRIPARTITE Remote Future tripartite number marking) 577 Table 14. Nungon tenses by number subsystem and behavior under negation 578 579 Following Aikhenvald & Dixon (1998), the Nungon tense system may be described as 580 dependent on polarity. Whether a clause is positive or negative affects the number of formal 581 tense inflections available for that clause. That is, two Nungon tenses—the Present tense and the 582 Near Future tense—only occur in positive-polarity clauses. To negate a clause framed in positive 583 polarity with a verbal inflected for Present tense, the verbal predicate must be reframed 584 in the Near Past tense. Examples are in (19) and (20).

585 19) Umum-o duo-ha-k. 586 snug-ADJ sleep-PRES.SG-3SG 587 S/he is sleeping snugly. 588 589 20) Ma=duo-wa-k, uyek it-ta-k. 590 NEG=sleep-NP.SG-3SG raw exist-PRES.SG-3SG 591 S/he is not sleeping, s/he is awake.

592 Note that in (20), if the Near Past form duo-wa-k ‘s/he slept’ were not negated, this could only 593 refer to sleeping that already took place some time before the speech act and after the morning of 594 the previous day. Here, the negated form ma=duo-wa-k is ambiguous: it could either mean that 595 sleeping did not already take place, or it could refer to sleeping that is not in progress at the time 596 of the speech act. 597 Similarly, when a positive-polarity clause with verbal predicate inflected for Near Future 598 tense is negated, the verb must be reframed in the negated Remote Future tense (this is also 599 identical to the negated Future Irrealis, since the Remote Future suffix -ma drops under 600 negation). This is shown in (21) and (22).

601 21) Unga, to-wangka-t. 602 today do-PROB.NF.SG-1SG 603 Today, I will do (it). 604 605 22) Ma=t-i-t, unga-i. 606 NEG=do-IRR.SG-1SG today-TOP 607 I won’t do (it), today.

608 Note that positive-polarity Remote Future t-i-t-ma ‘I will do (it)’ can strictly only apply to 609 ‘tomorrow’ and beyond.

610 Like the splits among pronouns, the number splits by tense suffix are unpredicted by the 611 Animacy Hierarchy. Tense would seem to be a grammatical category supremely oblivious to the 612 animacy of subject arguments. In fact, the two-axis schema in Figure 5 for representing number 613 splits within and between pronoun and noun paradigms is still insufficient to represent the 614 number splits between tense suffixes. Figure 8 gives a three-dimensional schema, with tense on 615 an additional axis. 616

617 618 619 Figure 8. Representation of all number splits in Nungon with three axes 620 621 There is no reason to suggest a hierarchy of tenses corresponding to number cross-linguistically. 622 As will be seen in the next section, similar number system splits depending on tense to those in 623 Nungon can be seen in other Finisterre Papuan languages; this may be due to the historical 624 development of tense inflections within the group. Such relationships between tense and number 625 system splits are as yet unattested for languages beyond the region. Interactions between number 626 and person, case, , and aspect are well-established for languages with single number 627 systems throughout their grammars (Aikhenvald & Dixon 1998: 195, Corbett 2000: 271–280). It 628 may well be that these grammatical systems are related to splits between number subsystems in 629 other languages with split number. These could then be represented using a three-dimensional 630 schema like that in Figure 8, albeit with a different category on the third axis.

631 7 Diachrony and split number

632 Some of the splits between number subsystems in Nungon are likely due to historical 633 developments. The most convincing data indicates historical reduction of number value

634 distinctions. A more speculative notion links the late development of future tense inflections, or 635 their historical instability, to tripartite number marking. 636 First, comparative data from related languages indicates that language history likely led 637 to the number split between first and second person basic and emphatic personal pronouns. The 638 Nungon basic pronouns non ‘1nsg’ and hon ‘2nsg’ probably originally designated only more 639 than two referents (i.e. ‘1pl’ and ‘2pl’), with original dual forms *not and *hot indicating exactly 640 two referents. The original plural forms then generalized into non-singulars, with the original 641 dual forms dropping out of use, a phenomenon well-attested on a larger scale from language 642 families as diverse as Semitic, Greek, and Slavic (references in Corbett 2000: 269n). 643 This hypothesis about the origin of the bipartite number system in the Nungon first and 644 second person basic pronouns is supported by language-internal and comparative data. First, the 645 first and second person plural Nungon emphatic pronouns noni and honi are identical to the first 646 and second person plural basic pronouns non and hon but for a final -i in the emphatic set. Since 647 the dual emphatic forms nori and hori also have this final -i, this could imply that if there were 648 original 1du and 2du basic pronouns, these would have the forms *not and *hot (/t/ becomes /r/ 649 intervocalically in Nungon morphophonology). Elsewhere in Nungon grammar, /r/ (or /t/) and /n/ 650 are the formants that indicate dual and plural, respectively, as in the Near Future and Remote 651 Future tense suffixes. 652 This conjecture also aligns with McElhanon’s 1970s comparative work, based on surveys 653 throughout the Finisterre-Huon region; McElhanon (1973: 20) reconstructs the proto-Finisterre- 654 Huon basic pronouns as including dual forms for all persons, with final segment -t in all forms. 655 Examination of the basic pronoun forms in two other Finisterre languages shows that there is a 656 tendency, apparently quite recent in some languages, toward reduction of number values 657 distinguished. Languages referenced in this section are all classed in the Finisterre language 658 group: Nukna, likely Nungon’s closest relative outside its dialect continuum (Author 2014b); Ma 659 Manda and Nek, both members of the Erap language family within Finisterre; Yopno, Nungon’s 660 immediate neighbor to the west, but different in many ways from Nungon; and Irumu, which 661 belongs to the Wantoat family within Finisterre. 662 Nek is spoken about 20 miles south of the Nungon-speaking region, on the other side of 663 the Saruwaged Mountains. Nek has a single set of personal pronouns that function both in 664 contexts where the Nungon basic pronouns are used and in contexts where the Nungon emphatic 665 pronouns are used (Linnasalo 1993: 10). The Nek personal pronouns operate according to the 666 maximal, tripartite number subsystem in that language, as seen in Table 15. 667 singular dual plural 1 nak nɨt nɨn 2 dɨk sɨt sɨn 3 en nɨsɨt nɨsɨn 668 Table 15. Nek personal pronouns (Linnasalo 1993:10) 669 670 There is some indication that the modern Nek personal pronouns are related to the 671 Nungon basic pronouns, and that Nek originally had a counterpart set of emphatic pronouns.

672 First, the basic/emphatic pronoun distinction is well-known for many Finisterre-Huon languages, 673 with McElhanon (1973: 21) even stating that all Finisterre-Huon languages documented by the 674 early 1970s showed this distinction. Second, Nek does maintain a distinction equivalent to that 675 between Nungon non=ton ‘our’ (basic pronoun with genitive clitic) and noni-win ‘our own’ 676 (emphatic pronoun with emphatic genitive suffix); in Nek, nɨn-dok ‘our’ contrasts with nɨne-lok 677 ‘our very own.’ It is likely that Nek nɨn is related to Nungon basic 1nsg pronoun non ‘we,’ and 678 Nek nɨne ‘we (emphatic)’ is related to Nungon emphatic 1pl pronoun noni. 679 While Nek pronouns may be more conservative than Nungon basic pronouns, with an 680 original tripartite number subsystem preserved for all persons, Linnasalo noted in the early 1990s 681 that the singular third person form en was in the process of generalizing to encompass dual and 682 plural values, writing that en was “being used for the third person, whatever the number” (1993: 683 10). Linnasalo also noted the curious forms of the Nek 3du and 3pl pronouns, which she 684 interpreted as clear formal fusions of the first and second person pronouns (1993: 10). 685 Nungon’s closest relative among non-mutually-intelligible Finisterre languages is Nukna 686 (see Author 2013 for an initial comparison). Contemporary Nukna first and second person basic 687 pronouns likely represent the middle stage of the historical transition in Nungon from a system 688 like that of Nek to the present-day Nungon basic pronouns. The Nukna basic pronouns are in 689 Table 16. 690 691 singular dual plural 1 nək nət nən 2 kək sən 3 wa 692 Table 16. Nukna basic personal pronouns (Taylor 2015: 31–32) 693 694 The three-way number system split in Nukna basic pronouns aligns perfectly with 695 Corbett’s assumptions about number system splits and the Animacy Hierarchy. At least the 696 number split between first and second persons, however, seems to be a recent and partially 697 incomplete development. For the first and second person pronouns, Nukna quite clearly is in the 698 process of generalizing original plural forms to encompass dual number as well: exactly the 699 historical process hypothesized for Nungon basic pronouns. In Nukna, this process is transparent 700 in several ways. First, the presence of a discrete /t/-final form for 1du, as in Nek, but the absence 701 of such a form for 2du, hints that generalization of the original plural form has likely occurred 702 for second person, but not yet for first person. 703 Further, the expected 2du form sət in fact still occurs in some contexts. Taylor (2013: 77– 704 79) supplies examples in Nukna showing that verbal agreement with a 2nsg basic pronoun 705 subject argument sən may be either dual or plural: without any grammatical relation-marking 706 enclitics, the original plural form *sən has fully generalized to encompass the dual. But when the 707 Comitative enclitic =yot occurs on a basic pronoun with 2du reference, sən is not used; instead, 708 the expected, older form *sət resurfaces: sət=yot is used for ‘with you two,’ and *sən=yot is in

709 fact unacceptable. In most villages, the pronoun keeps the form sən before the genitive enclitic 710 =Cən even with dual reference, yielding sən=dən ‘of you two.’ But in the Nukna dialect of three 711 villages, the form is sət=(t)ən, reflecting the older form *sət. 712 While Taylor (2015: 31–32) classes wa as a third person personal pronoun, it must be 713 noted that this is homophonous with what he lists as a demonstrative meaning ‘that’ (2015: 53), 714 cognate with Nungon wo ‘that.’ Thus, Nukna could alternatively be said to lack a dedicated third 715 person personal pronoun, with the distal demonstrative functioning as such a pronoun in 716 discourse. As Dixon (2010b: 190) writes, this recruitment of the distal demonstrative as a de 717 facto third person pronoun is common cross-linguistically. Indeed, the cognate distal 718 demonstrative wa ‘that’ is also used in lieu of a dedicated third person pronoun in the Finisterre 719 language Ma Manda (Pennington 2016: 258). 720 The third person pronouns in Nek and Nungon similarly diverge from McElhanon’s 721 hypothesized proto-Finisterre-Huon basic third person pronouns, although in neither language is 722 the distal demonstrative coopted. The Nek 3du and 3pl pronouns nɨsɨt and nɨsɨn look like 723 combinations of 1du nɨt and 2du sɨt, and 1pl nɨn and 2pl sɨn; this formal ‘constructed person’ (where first 724 plus second apparently indicates third) could be an innovation special to Nek. Nungon and Nek are like 725 Nukna and Ma Manda in that there is a single third person pronoun form used for all numbers (or trend 726 toward finalizing this, in Nek). In Nungon and Nek, third person forms seem to be characterized by 727 more drastic leveling—the simultaneous application of a singular form to both dual and plural 728 referents—than first and second persons, and in both languages it is the original singular third 729 person form that generalizes to encompass dual and plural reference. As noted above, first and 730 second persons are known to pattern together in opposition to third person in pronoun sets in 731 some languages (Dixon 2010b: 190). For instance, in Dyirbal, first and second person pronouns 732 are case-marked on a nominative-accusative basis, while third person pronouns and all nouns are 733 case-marked on an absolutive-ergative basis (Dixon 1994: 86). The third person basic pronouns 734 in Nek and Nungon pattern with non-human nouns—and, indeed, unpossessed human nouns—in 735 never marking number. 736 It is reasonably probable that the Nungon first and second person plural basic pronouns 737 have generalized into non-singular pronouns, and that the same occurred with the Nukna second 738 person plural basic pronoun. Language-wide loss of the dual number value is documented in 739 languages such as Hebrew and Arabic, Greek, and varieties of Slavic (Corbett 2000: 269). The 740 phenomenon in Nungon and Nukna pronouns could be a microcosm of such larger-scale loss. 741 The direction of formal generalization is not a given here. As McElhanon (1973: 20) points out, 742 some of Nek’s close relatives in the Erap family have apparently generalized the original first 743 and second person dual pronoun forms, rather than the original plural forms, into non-singulars. 744 This is evident in Uri (McElhanon 1973: 20) and in Ma Manda (Pennington 2016: 258). Use of 745 original dual forms to index non-singular number is best attested in modern Arabic varieties 746 (Blanc 1970) and in other languages such as Russian for specific lexical items that usually occur 747 in pairs (such as ‘eyes’). (Also see Palancar 2013 on Otomi.) In these other languages, however, 748 an original dual number value and original dual forms have otherwise been largely lost (Corbett 749 2000: 269, 286). In the Finisterre languages, apparent constrictions of number systems are

750 restricted to certain parts of the grammar (rather than certain lexical items), leading to number 751 system splits. 752 In sum, comparative evidence from related languages, including Nek and Nukna, points 753 to an original tripartite number subsystem in the first and second persons, and likely in the third 754 as well, of the Nungon basic free pronouns. If this were still the case, plotting the basic pronouns 755 with nouns along the Animacy Hierarchy would yield no violations of the hierarchy. Historical 756 generalizations—of the original plural form into a non-singular for first and second persons, and 757 of the original singular form into the sole form, not distinguishing number, for the third person— 758 are proposed to have led to the current situation, with violations of two major stipulations of the 759 hierarchy. The further questions, of course, are why the original tripartite system has constricted 760 to bipartite or null among basic pronouns in disparate languages within the Finisterre-Huon 761 group, and why the third person seems especially prone to reduction in number value 762 distinctions, and even to elimination and replacement with the distal demonstrative ‘that.’ 763 Within the pronouns, the other number split is among bound pronouns: between the 764 pertensive markers and verbal subject suffixes, with a tripartite number subsystem, and the 765 verbal object prefixes, with a bipartite one. This split likely relates to historical change, canonical 766 constituent order, and the locus of formal number marking in free pronouns, but originated deep 767 in Nungon language prehistory. 768 Comparative data indicates that the Nungon object prefixes are formally similar to object 769 prefixes found throughout the Finisterre-Huon group. Indeed, Suter (2012) shows that 770 pronominal prefixes indexing object person and number are found in most Finisterre-Huon 771 Papuan languages. He claims that rather than developing separately in each language, these 772 prefixes are of ancient provenance, reaching back to proto-Finisterre-Huon, and possibly even 773 beyond, since languages in the New Guinea Highlands have verbs with apparently cognate object 774 prefixes (2012: 31).6 775 It is reasonable to suppose that because all these languages have verb-final constituent 776 order, with object arguments canonically immediately before the verb, that the object prefixes 777 originated as free pronouns referring to object arguments. In modern Nungon, free basic 778 pronouns are much more common than emphatic pronouns with object reference; if this were the 779 case in proto-Finisterre-Huon, it would have been the basic pronouns that proclitized, then 780 became prefixed, to certain transitive verbs. McElhanon’s reconstruction of the proto-Finisterre- 781 Huon basic personal pronouns (1973: 20) is shown in Table 17. 782 783 784 785 786

6 Suter (2012: 43) shows that the Finisterre-Huon group is divided as to which number subsystem object prefixes follow: in the Huon Peninsula group, the object prefixes of most languages operate according to the tripartite subsystem, while those of all documented languages in the Finisterre group operate according to the bipartite system.

singular dual plural 1 *nak *nit *nin 2 *gak *ŋgit *ŋgin 3 *yak *yit *yin 787 Table 17. McElhanon’s reconstruction of proto-Finisterre-Huon basic pronouns 788 789 In McElhanon’s reconstruction, as in modern Nek pronouns, number distinctions between dual 790 and plural are marked through the final consonant of the forms. If these were prefixed to a verb, 791 losing their final consonant, all formal distinction between dual and plural would be lost. Here, 792 then, constituent ordering has apparently interacted with morphology (the precise point in the 793 pronoun where number is formally distinguished) to create a situation where verbal object 794 prefixes operate according to a different number subsystem than verbal subject suffixes. 795 Tense markers on verbs are an unlikely place to find a split between number subsystems. 796 Again, consulting other Finisterre languages shows that Nungon is not alone in number splits 797 between tenses. Tense markers in Nek (Linnasalo 2014), Yopno (Reed 2000), Ma Manda 798 (Pennington 2014), and Irumu (Ross Webb, p.c.) are fused with subject number indexation; as in 799 Nungon, this number is generally superfluous in that subject number is also indexed in a final 800 person/number suffix. The number subsystems according to which the tense/number suffixes 801 operate in these languages are shown in Table 18. 802 Remote Past Near Past Present Near Future Remote Future Nungon NULL BIPARTITE TRIPARTITE Yopno NULL BIPARTITE TRIPARTITE Nek BIPARTITE NULL BIPARTITE TRIPARTITE Ma Manda BIPARTITE TRIPARTITE Irumu NULL BIPARTITE TRIPARTITE BIPARTITE 803 Table 18. Number subsystems by tenses in related languages 804 805 In Table 18, a few patterns in number subsystem distribution are evident: the tripartite number 806 subsystem is only found in future tenses, the present tense always uses a bipartite subsystem, and 807 the null subsystem only occurs in the past tenses. 808 The historical explanation suggested here for the number value splits between tenses 809 involves a trajectory that is necessarily the opposite of that proposed for the personal pronouns at 810 the beginning of this section. The original Nungon basic personal pronouns were proposed to 811 have operated with a maximal tripartite number subsystem. The first and second person basic 812 pronouns then lost a formal distinction between dual and plural number, while the third person 813 basic pronoun lost all formal number distinctions, retaining what was likely the original singular 814 form. If Nek and Nukna yield any indication of how the process unfolded in Nungon, number 815 distinctions were lost beginning with the third person (as in Nek), followed by the second person 816 (as in Nukna), and ending with the first person. With the tenses, however, the modern forms with

817 the maximal number value distinctions are taken here to be the newest forms to develop, 818 followed in newness by the forms with the medial number value distinctions. The tense form 819 with the minimal, null, number value distinction is taken as the oldest. It is as yet unclear which 820 changes occurred at a proto-Finisterre stage, and which in the individual languages.7 821 In many familiar languages, future tenses were late to develop, and bear clear 822 morphological traces of their origins in auxiliaries of intention or prediction (Comrie 1985: 44– 823 45, Dahl 1985, Bybee, Perkins & Pagliuca 1994: 243–280). Within the Nungon dialect 824 continuum, only the form of the Remote Future inflection varies significantly from the northern 825 to the southern half of the continuum. The Nungon Remote Future form is related to the 826 equivalent of the Future Irrealis in other dialects, while the Remote Future form in those dialects 827 is nonexistent in Nungon. The mutability of future forms across Finisterre languages could be 828 due to their relatively recent development.8 It is suggested here that the future tenses could show 829 tripartite number marking on their suffixes because these suffixes were originally word-final 830 markers of number on reduced, dependent verb forms functioning together with an auxiliary verb 831 bearing full subject person/number inflection. Indeed, a vestige of such a number- (but not 832 person-) marked verb form still exists in Nungon and some related languages. 833 In modern Nungon and in at least two other Finisterre languages, Awara and Tayatuk 834 (Quigley 2014, Guérin 2015), there exists a minor independent verbal inflection that features 835 only the verb root plus a suffix fused with number, but no final subject person/number suffix. 836 The form has general future or modal meaning in each language. In Nungon, I have called this 837 form the Probable (Author 2014a, 2014c). Since the Nungon Probable lacks specification of 838 subject person, such specification must be either supplied explicitly elsewhere in the clause, or 839 recovered from context. The Probable suffix forms are in Table 19. 840 singular dual plural -wang -rang -nang 841 Table 19. Nungon Probable suffixes 842 843 An example of the Probable in use is in (23).

844 23) Mee, ga-mo-wang. 845 later 2SG.O-give-PROB.SG 846 (I) will/would give (it) to you later.

7 Preliminary examination of grammars indicates that at least the Nungon Future Irrealis (formal basis for the Remote Future tense) has cognates in a number of other languages. 8 The most archaic tense form in Nungon and its relatives could be the Remote Past form; this form seems to vary the least across all the languages examined. It is possible that such a form originally coexisted with another, tense- neutral, form, the origin of today’s Immediate Imperative, Counterfactual, and switch-reference different-subject inflections (see Table 5). See Author (2015) for more on this tense-neutral form, which would have been similarly “chameleon-like” to the Indo-European Injunctive (Kiparsky 1968, 2005).

847 In (23), there is no formal indication of the person of the subject argument of ‘give’; this could 848 be first, second, or third person depending on speech act context. 849 The Probable form is also formally related to the Desiderative form, which comprises the 850 Probable, inflected for appropriate number value, with an unchanging suffix -na. This form may 851 stand alone as an attributive adjective indicating intentionality (with subject arguments capable 852 of intention) or imminent action, or both. But the Desiderative is most often followed by the 853 inflected verb to- ‘do,’ with the same indications of intentionality and/or imminence as when it 854 occurs alone. An example with a weather verb is in (24).

855 24) Iso-wang-na to-go-k. 856 dawn-PROB.SG-DESID do-RP-3SG 857 It was about to dawn.

858 The Near Future tense forms, as seen in Table 12, are the only tense suffixes that are 859 disyllabic. The first syllable of these forms varies according to three number values and is 860 identical to the Probable suffixes, while the second syllable is the non-varying /ka/. The Probable 861 form always relates to unrealized future events. It is quite possible that the Near Future forms 862 originated as auxiliary constructions with a number-marked Probable verb form followed by a 863 ‘do’ or ‘be’-type auxiliary. In fact, in the Yawan village dialect of Nungon, the Near Future 864 suffix takes the form -wangta, -rangta, -nangta, rather than -wangka, -rangka, -nangka of the 865 Towet dialect given in Table 12. This makes the Yawan dialect Near Future forms homophonous 866 with the Probable followed by the verb to- ‘do’ in Present tense, except that the Present tense 867 inflections of to- ‘do’ have a long vowel /aa/. The speculative origin of the Near Future as an 868 auxiliary verb construction with number-marked lexical verb is shown in Figure 9. 869 Auxiliary construction Fusion *ongo-rang ta-a-mok go-PROB.DU do-PRES-1DU ongo-rang-ta-(a-)mok ongo-rang-ka-mok ‘We two are about to go’

postulated retention of loss of primary stress on /ta-a-/ assimilation of primary stress on both resulting in accompanying loss consonant /t/ to components of construction of vowel length preceding velar nasal

(present-day Yawan dialect (present-day Towet

Near Future) dialect Near Future) 870 Figure 9. Hypothesized development of the Nungon Near Future tense form 871 872 Development of the Near Future tense from an auxiliary construction, with the first, 873 lexical, verb marked for number but not person, is one potential explanation for double-marking 874 of subject number according to the maximal tripartite system, in that tense. Unfortunately, such a 875 ready explanation is not presently available for the Nungon Remote Future tense.

876 As for the bipartite number subsystem operating in the Present and Near Past tenses, 877 there are several possible explanations. It is likely that these forms in Nungon are older than the 878 future forms, and historical clues are shrouded after centuries of phonological change. For both 879 this and the Remote Future provenance mystery, the coincidence of tense neutralization under 880 negation with shared number subsystems may be helpful in further probing time-worn 881 morphology. 882 In sum, history very likely explains the number split among the Nungon free pronouns: 883 the postulated original tripartite system for the basic pronouns was reduced to a bipartite system 884 for first and second persons and a null system for third person. Elimination of the dual number 885 value throughout a language’s grammar is well-attested from familiar languages. With the tenses, 886 it may be speculated that tripartite number marking reflects historical fusion of the components 887 of special auxiliary constructions, but there is no firm evidence for this, especially with the 888 Nungon Remote Future. The similarity in distribution of number subsystems across the tenses in 889 related languages also suggests that the splits date back to proto-languages. 890 There is as yet no reason to propose that Nungon, and other Finisterre and non-Finisterre 891 Papuan languages with split number, originally exhibited tripartite number marking throughout 892 their grammars. The number splits seen in these languages are embedded deep in inflectional 893 morphology and pronoun forms.

894 8 Conclusion

895 This has been the first in-depth exposition of an entire split number system. Nungon shows that 896 forms, as well as bound pronouns, can operate according to different number 897 subsystems than non-reflexive free pronouns. Further, number system splits may occur both 898 among controllers (nouns or pronouns) and, independent of the controller system, among targets 899 (modifiers and verbs that index controller number). The places in Nungon grammar in which the 900 three number subsystems function are summarized in Table 20. 901 number subsystem tripartite bipartite null free pronouns emphatic 1st and 2nd person 3rd person basic pronouns basic pronouns pronoun

nouns prototypically- — prototypically-non- human nouns human nouns

bound pronouns subject suffixes, object prefixes — pertensive markers tenses Remote Future, Present, Near Past Remote Past Near Future 902 Table 20. Summary of Nungon number system splits

903 904 If a number system split occurs within a single Nungon paradigm, this accords with the 905 Most Values at the Top conjecture of the Animacy Hierarchy—for controller number. The first 906 and second person basic pronouns use the bipartite number subsystem, while the third person 907 pronoun has the null subsystem. Similarly, among nouns, kin terms and non-kin prototypically- 908 human nouns feature the tripartite subsystem, while all other nouns use the null subsystem. But 909 when all paradigms are considered together against the Animacy Hierarchy, the Most Values at 910 the Top stipulation fails. 911 Splits between pronoun sets, as between the basic and emphatic free pronouns, and 912 between the subject and pertensive bound pronouns, and the object bound pronouns, are 913 unpredicted by the Animacy Hierarchy. These number system splits must be accounted for in 914 other ways, along a dimension unrelated to the Animacy Hierarchy itself (Figure 5). Like these, 915 number splits within the tense suffixes are unrelated to the Animacy Hierarchy. But unlike the 916 splits between pronoun sets, those between tenses occur in a category that is not even represented 917 on the Animacy Hierarchy, since the splits relate solely to tense, independent of subject 918 reference. The schema in Figure 7 showed the multidimensionality of Nungon number system 919 splits. 920 The most compelling explanation for the split between the emphatic and basic pronoun 921 sets is historical. Assuming that the Nungon basic pronouns originally all marked number 922 according to a tripartite subsystem, different number values were lost for different persons, in 923 accordance with the Animacy Hierarchy as formulated by Corbett (2000: 56). That is, in the third 924 person, the original singular form may have generalized to encompass dual and plural reference, 925 while in the first and second persons, it was the plural form that generalized into a non-singular, 926 with the original singular form remaining in opposition to this form. The basic pronouns are 927 functionally unmarked relative to the emphatic pronouns, and are also more frequent in the texts 928 corpus than emphatic pronouns. But frequency and functional are not the only 929 factors in reduction of number values; both first and second person basic pronouns are more 930 frequent in texts than the third person basic pronoun, which shows the most reduction. With the 931 impending compilation of a Nungon natural speech corpus composed largely of dialogue, 932 pronoun frequency counts in third-person narratives will be compared to frequency counts in 933 dialogues. Frequency and functional markedness may have played roles in a historical shift in the 934 basic pronouns that did not also occur in the emphatic pronouns. 935 Discussion of the places in Nungon grammar where number system splits occur 936 highlights the interrelatedness of four different grammatical categories in Nungon: number, 937 person, polarity, and tense. Most of the dependencies between Nungon grammatical systems are 938 anticipated by Aikhenvald & Dixon’s hierarchy (1998, 2011), but the dependency between 939 number and tense is as yet undescribed. Here, following Aikhenvald & Dixon, the expression 940 X > Y should be read ‘Y is dependent on X.’ 941 942 1. Person > number, number > person. Person is dependent on number in the verbal subject 943 suffixes on all verbs inflected for subject person: In the dual and plural numbers, there are fewer

944 formal person distinctions in these suffixes. But number is dependent on person in the basic 945 pronouns, where the first and second persons indicate different number values than the third 946 person. 947 948 2. Polarity > tense. In Nungon, the number of formal tense distinctions available depends on 949 polarity, with five tense distinctions in positive polarity and only three under negation. This does 950 not also number marking in the tense suffixes, however, since neutralization under 951 negation coincides with number subsystem splits in positive polarity (see Table 14). 952 953 3. Tense > number. In the tense suffixes, the number of number values distinguished in 954 indexation of the subject argument depends on the choice of tense. Aikhenvald & Dixon’s 955 hierarchy (2011: 195) predicts that if a dependency were found in a language between tense and 956 number, it should be number that is dependent on tense, not the other way around. But they cite 957 only one related example, itself a complex dependency, where person and number marking 958 depend on combined polarity, tense and evidentiality values (in Udmurt, 2011: 193). A simple 959 dependency between number and tense like that of Nungon and related languages is as yet 960 undescribed in these terms. 961 962 This has been the first case study of where number subsystem splits occur throughout the 963 entire grammar of a single language. The exposition showed that the Animacy Hierarchy is 964 useful in explaining some splits, but not all. The three number subsystems are all widespread 965 throughout Nungon grammar, including in unlikely places such as tense markers, where different 966 number subsystems function in different tenses. The splits between tenses further correspond to 967 neutralization of some tense distinctions under negation. Although most Finisterre Papuan 968 languages likely exhibit similar complexity in number splits, it remains to be seen whether such 969 multi-dimensionality is in fact the norm in split number systems cross-linguistically.

970 Abbreviations

971 1, 2, 3 - person ADJ - adjectivizer ASSOC - associative AUTOREFL - autoreflexive 972 COLL - collective COMIT - comitative COMPL - completive CONTR - counterfactual 973 DEP - dependent DESID - desiderative DS - different subject DU - dual EMPH - emphatic 974 FOC - GEN - genitive IRR - future irrealis LOC - locative MV - medial verb 975 NEG - negator NF - near future NP - near past NSG - non-singular O - object PL - plural 976 POSS - pertensive PRES - present PRO - personal pronoun PROB - probable 977 RF - remote future RP - remote past RSTR - restrictive, durative SG - singular 978 TOP - topicalizer

979 References

980 Aikhenvald, Alexandra Y. and R. M. W. Dixon. 1998. Dependencies between grammatical 981 systems. Language 74: 56–80. 982 Aikhenvald, Alexandra Y. and R. M. W. Dixon. 2011. Language at Large: Essays on Syntax and 983 Semantics. Leiden: Brill.

984 Blanc, Haim. 1970. Dual and pseudo-dual in the Arabic dialects. Language 46: 42–57. 985 Bybee, Joan, Revere Perkins, & William Pagliuca. 1994. The Evolution of Grammar: Tense, 986 Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. 987 Childs, G. Tucker. 2005. The S-AUX-O-V syntagm in the Atlantic languages. Studies in African 988 Linguistics 33, 2: 1–41. 989 Claassen, Oren R. and Kenneth A. McElhanon. 1970. Languages of the Finisterre Range - New 990 Guinea. Papers in New Guinea Linguistics 11, 45–78. 991 Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press. 992 Comrie, Bernard. 1989. Language Universals and Linguistic Typology, Second Edition. Chicago: 993 University of Chicago Press. 994 Corbett, Greville G. 2000. Number. Cambridge: Cambridge University Press. 995 Croft, William. 2003. Typology and Universals: Second Edition. Cambridge: Cambridge 996 University Press. 997 Dahl, Östen. 1985. Tense and Aspect Systems. Oxford: Blackwell. 998 Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press. 999 Dixon, R. M. W. 2010a. Basic Linguistic Theory, Volume 1: Methodology. Oxford: Oxford 1000 University Press. 1001 Dixon, R. M. W. 2010b. Basic Linguistic Theory, Volume 2: Grammatical Topics. Oxford: 1002 Oxford University Press. 1003 Dixon, R. M. W. 2012. Basic Linguistic Theory, Volume 3: Further Grammatical Topics. 1004 Oxford: Oxford University Press. 1005 Elšik, Viktor & Yaron Matras. 2006. Markedness and Language Change: The Romani Sample. 1006 Berlin: Mouton de Gruyter. 1007 Foley, William A. 1986. The Papuan Languages of New Guinea. Cambridge: Cambridge 1008 University Press. 1009 Guérin, Valérie. 2015. Imperatives and commands in Tayatuk. Presented at the International 1010 Workshop on Commands, Cairns: James Cook University. 1011 Heine, Bernd. 1982. African systems. In Hansjakob Seiler & Christian Lehmann 1012 (eds.), Apprehension: das sprachliche Erfassen von Gegenständen, i: Bereich und Ordnung der 1013 Phänomene. Tübingen: Narr. 189–216. 1014 Kiparsky, Paul. 1968. Tense and mood in Indo-European syntax. Foundations of Language 1015 4(1):30–57. 1016 Kiparsky, Paul. 2005. The Vedic injunctive: historical and synchronic implications. In Rajendra 1017 Singh & Tanmoy Bhattacharya (eds.), The Yearbook of South Asian Languages 2005. New 1018 Delhi: Sage Publications. 219–235. 1019 Linnasalo, Katri. 1993. Nek Grammar Essentials. Ms. Ukarumpa: Summer Institute of 1020 Linguistics. 1021 Linnasalo, Katri. 2014. Non-spatial setting in Nek. In Hannah Sarvasy (ed.), Non-Spatial Setting 1022 in Finisterre-Huon Languages. Sprachtypologie und Universalienforschung 67(3): 297–326. 1023 Lockwood, Hunter T. & Monica Macaulay. 2012. Prominence hierarchies. Language and 1024 Linguistics Compass 6/7: 431–446.

1025 McElhanon, Kenneth A. 1973. Towards a Typology of the Finisterre-Huon Languages. 1026 Canberra: Pacific Linguistics. 1027 Moravcsik, Edith. 2003. A semantic analysis of associative . Studies in Language 27(3): 1028 469–503. 1029 Palancar, Enrique. 2013. The evolution of number in Otomi: the many faces of the dual. Studies 1030 in Language 37(1): 94–143. 1031 Pennington, Ryan. 2014. Non-spatial setting in Ma Manda. In Hannah Sarvasy (ed.), Non-Spatial 1032 Setting in Finisterre-Huon Languages. Sprachtypologie und Universalienforschung 67(3): 327– 1033 364. 1034 Pennington, Ryan. 2016. A grammar of Ma Manda. Cairns: James Cook University PhD thesis. 1035 Quigley, Susan. 2014. Non-spatial setting in Awara. In Hannah Sarvasy (ed.), Non-Spatial 1036 Setting in Finisterre-Huon Languages. Sprachtypologie und Universalienforschung 67(3): 365– 1037 393. 1038 Reed, Wes. 2000. Yopno Grammar Essentials. Ms. Ukarumpa: Summer Institute of Linguistics. 1039 Roberts, John R. 1997. Switch-reference in Papua New Guinea: a preliminary survey. In Andrew 1040 Pawley (ed.), Papers in Papuan Linguistics 3. Canberra: Pacific Linguistics. 101–241. 1041 Silverstein, Michael. 1976. Hierarchy of features and ergativity. In R. M. W. Dixon (ed.), 1042 Grammatical Categories in Australian Languages. Canberra: Australian Institute of Aboriginal 1043 Studies. 112–171. 1044 Smith-Stark, T. Cedric. 1974. The plurality split. Papers from the Annual Regional Meeting of 1045 the Chicago Linguistic Society 10: 657–671. 1046 Suter, Edgar. 2012. Verbs with pronominal object prefixes in Finisterre-Huon languages. In 1047 Harald Hammarström & Wilco van den Heuvel (eds.), History, Contact, and Classification of 1048 Papuan Languages, Part One. Special issue of Language and Linguistics in Melanesia. 23–58. 1049 Taylor, Matthew. 2013. Nukna Grammar Essentials: In Progress. Ms. Ukarumpa: Summer 1050 Institute of Linguistics. 1051 Taylor, Matthew. 2015. Nukna Grammar Sketch. Ukarumpa: Summer Institute of Linguistics.