Irrealis Is Real 1 Introduction

Home , Irrealis mood, Paamese language, Realis mood

Irrealis is real

1 Abstract

2 The question whether irrealis is a meaningful concept in cross-linguistic comparison has been 3 the subject of longstanding controversy. In this paper, we will argue that the semantic domain 4 of irreality is split into two domains—the possible and the counterfactual—and that an “irrealis” 5 marker in a given language may refer to only one of these domains or to both. A significant part 6 of the cross-linguistic variation in what is referred to by the term irrealis can be traced back to this 7 distinction. Other factors which obscure the realis/irrealis divide include functional subdivisions 8 of the irrealis domain and paradigmatic competition within the TAM system of a language. We 9 conclude that irrealis is a cross-linguistically meaningful notion.

10 1 Introduction

11 1.1 Overview

12 Many languages have been described as systematically distinguishing between utterances relating 13 to actual events, and utterances referring to future, hypothetical or counterfactual scenarios. This 14 phenomenon is usually characterized as a distinction between realis and irrealis mood. This is 15 illustrated by the following examples from Nanti (Arawak) (Michael, 2014: 254):

16 (1) a. O=pok-∅-i maika 3.nonm.sbj=come-ipfv-real.i now 17 “She is coming now.” 18 b. o=N-pok-∅-e kamani 3.nonm.sbj=irr-come-ipfv-irr.i tomorrow 19 “She will come tomorrow.”¹

20 However, this practice has been dismissed for cross-linguistic research by some. The main points 21 of criticism boil down to the following arguments: 22 1. The notion of irrealis is conceptually unappealing. 23 2. Few languages have a binary distinction between realis and irrealis mood. Compare Bybee 24 (1998: 265):

25 For any given language, there are several grams that mark off portions of the con- 26 ceptual space for situations that are not asserted to exist, or if there is a highly 27 generalized gram, it does not cover all “irrealis” situations […]

28 3. The functions of markers that have been labeled as irrealis are too inconsistent cross-linguistically 29 to constitute a useful category. We will discuss this argument in detail in section 4.

¹agr – agreement; ag – agentive; al – alienable; aux – auxiliary; ben – benefactive; cf – counterfactual; cl(1/2/3) – possessive classifier (class); comp – complementizer; cond – conditional; ctf – counterfactual; dem – demonstrative; dist – distal (TAM); dl – dual; du – dual; excl – exclusive; fut – future; habit – habitual; impf – imperfective; im – imminent; incl – inclusive; ipfv – imperfective; ipf – imperfective; irr – irrealis; i – i-class verb; neg1 – negation one (in Nafsan); neg2 – negation two; neg – negation; nim – non-imminent; nonm – non-masculine; nonsg – non-singular; pc – paucal; pl – plural; poss – possessive; pot – potential; prf – perfect; psp – prospective; real – realis; redup – reduplication; rel – relative clause complementizer; res – resultative; sbj – subject; sg – singular; ss – same subject/reference marker (Watahomigie et al. , 1982); sub – subordinator; tent – tentative; tma – tense, aspect, mood; trans – transitive; v – epenthetic vowel;

1 30 We believe that these points do not undermine the validity of the concept of irrealis. Instead, we 31 will argue that they can largely be resolved by a better theoretical understanding of the concept 32 and by more systematic typological and descriptive research. In particular, we will discuss the 33 following factors as responsible for a substantial portion of inconsistencies in how irrealis has 34 been described: 35 1. The two terms “realis” and “irrealis” can be given conceptually meaningful content, but the 36 lack of a theoretical basis has so far prevented this (section 2). 37 2. There is a binary distinction between realis and irrealis, but both are further divided into 38 temporal-modal and functional subdomains. Languages differ in how they carve up this 39 semantic space (section 3). 40 3. Mismatches between language-specific descriptions can be caused by the following factors: 41 (a) In some languages, markers that only refer to a sub-domain of irrealis are labeled as 42 irrealis. 43 (b) It is often assumed that realis, in particular, is expressed by a zero-morpheme orbya 44 portmanteau morpheme, but in some cases, the corresponding expression will simply 45 not encode any specific temporal-modal reference at all. This explains why forms that 46 are thought to encode realis may show up in non-realis contexts. 47 (c) Certain functions typically associated with irrealis may be encoded by more specific 48 markers in some languages and block the use of irrealis, such as specialized imperative, 49 prohibitive or timitive expressions. 50 These factors are discussed in section 4.

51 1.2 Historical background

52 The opposition between realis and irrealis moods has been observed early, and in a wide variety 53 from languages, especially from Oceania and the Americas. Possibly the first instance of the term 54 irrealis comes from Sapir’s description of Southern Paiute (Sapir, 1930: 168), where he states about 55 a particular suffix:

56 This element indicates that the activity expressed by the verb is unreal, i. e.either 57 merely potential or contrary to fact (potential in past time).

58 Already in this short quote, there are two very important elements present that we will high- 59 light throughout this paper: 1) The subdivision of irreality into the potential and the counterfactual 60 and 2) the connection between these modal domains and time. 61 Another early reference to the irrealis distinction comes from Dempwolff (1939), who uses the 62 terms modus realis and modus imaginativus—the latter has been translated as irrealis in Dempwolff 63 et al. (2005): The Oceanic language Yabem (or Jabêm) of Papua New Guinea has a paradigm ofverb 64 inflections that simultaneously encode person-number features of the subject, and a distinction 65 between realis mood, pertaining to assertions of the past and present, and irrealis mood, pertaining 66 to directives, assertions about the future and epistemic possibilities. We will see that this exact 67 situation is very widespread among Oceanic languages. 68 Capell (1971) acknowledges the irrealis distinction as an important feature of at least some 69 Austronesian languages of Papua New Guinea:

70 A further feature of [Austronesian languages of New Guinea with VO order] worthy 71 of attention is the general presence of a Realis- Irrealis distinction in the verbal system, 72 i. e. a basic distinction between actions which are regarded as actually occurring and 73 actions which are merely thought about. (Capell, 1971: 288)

2 74 In subsequent years, the term irrealis is also mentioned in Bickerton (1975), Johnston (1980), 75 Dixon (1980), Givón (1982), Chung & Timberlake (1985), Foley (1986) and Blewett, S.C. (1991) (also 76 see references in Mithun 1995). Before the 1990s, however, the irrealis distinction was primarily 77 used descriptively, with respect to individual, language-specific phenomena, and without a discus- 78 sion of its theoretical status in semantics, cognition, or typology. More systematic treatments only 79 emerged in the 1990s, starting with Roberts (1990) and Bugenhagen (1993), which we will discuss 80 in more detail in the following sections. 81 Around the same time, Trask (1993: 147) comes to the following assessment:

82 [Irrealis:] A label often applied in a somewhat ad hoc manner to some distinctive gram- 83 matical form, most often a verbal inflection, occurring in some particular language and 84 having some kind of connection with unreality. Palmer (1986) recommends that this 85 term should be avoided in linguistic theory on the ground that it corresponds to no 86 consistent linguistic content.

87 Bybee et al. (1994) and Bybee (1998) expand on this criticism of the category, culminating in the 88 three main objections listed in the introduction. In the context of the publications of Bybee et al. 89 (1994) and Bybee (1998), there were a number of language-specific and more general descriptions 90 that on the one hand defended the use of the label irrealis for the purposes of language-specific 91 description, but, on the other hand, presented linguistic data that sometimes appeared to confirm 92 the concerns of Bybee and others. Thus, Givón (1994) tries to relate irreality to the better-known 93 categories of epistemic and deontic modality, through data from Romance and Bantu languages, 94 even though neither have been described as basing their TAM systems around the realis/irrealis 95 distinction. 96 In 1995, Joan Bybee and Suzanne Fleischman edited a book on modality (Bybee & Fleischman, 97 1995), which includes discussions of irrealis by Chafe (1995) and Mithun (1995). Both contributions 98 come to the conclusion that, while there is considerable cross-linguistic variation in how irrealis 99 markers are distributed, there also appears to be a common, relatively stable core of functions 100 associated with this category. 101 In the same special edition Anthropological Linguistics that features Bybee (1998), there are also 102 descriptions by Kinkade (1998) on Upper Chehalis (Salish), Callaghan (1998) on Lake Miwok (Yok- 103 Utian) and Martin (1998) on Mocho (Mayan). It is not clear to us what prompted the selection 104 of these particular languages, since they had never been argued to systematically encode the re- 105 alis/irrealis distinction. It is therefore not surprising that a picture of bewildering inconsistency 106 emerges from comparing the different accounts, which Bybee (1998) takes as evidence that the 107 irrealis is not a typologically meaningful concept. Several later accounts tend to agree with this 108 assessment, including de Haan (2012) and Cristofaro (2012). 109 This criticism notwithstanding, researchers working on individual languages or groups oflan- 110 guages have maintained that the realis/irrealis distinction is a meaningful one, many of which 111 come from the context of Oceanic. These studies include Elliott (2000); Verstraete (2005); McGre- 112 gor & Wagner (2006); Van Gijn & Gipper (2009); Barbour (2011); Exter (2012); Cleary-Kemp (2014); 113 Michael (2014); Lichtenberk (2016). We weigh in on this debate by providing a new theoretical ba- 114 sis for the conceptual distinction, by applying the corresponding definition to individual in-depth 115 case studies, and by correlating our findings with a larger-scale study on cross-linguistic variation, 116 with a focus on the Oceanic context.

3 117 1.3 Data and methodology

118 For this paper, we draw on a wide range of empirical resources. We did several in-depth case 119 studies on seven Oceanic languages based on substantial corpus data. We also carried out targeted 120 elicitations on six of those languages to test specific hypotheses. These studies have been described 121 in depth in several publications, including [redacted], [redacted], [redacted], [redacted], [redacted], 122 [redacted]. We also compiled a database of Oceanic TAM systems based on existing descriptions, 123 on a total of 76 languages, although most factors could only be determined with certainty for a 124 subset of these. We comment on the sources of our examples and statistics throughout the article. 125 While our own empirical work focuses on Oceanic, our analyses also appear compatible with 126 observations on other languages as reported in the wider literature.

127 2 Irrealis in Branching Time

128 The realis/irrealis distinction is based on an inextricable link between modality and tense.The 129 intimate relation between these two domains has been discussed extensively in the literature on 130 tense semantics (Condoravdi 2002; Iatridou 2000, [redacted]). Moreover, the distinction between 131 realis and irrealis moods relies heavily on an asymmetry between the past and present on the one 132 hand and the future on the other hand. As Comrie (1985: 51) already notes, realis/irrealis systems 133 are often described in terms of future/non-future reference, even though irrealis markers arenot 134 restricted to the future, but can also refer to the possible or counterfactual past and present (also 135 compare Bhat 1999: 131). 136 The asymmetry between the future and the non-future is not only often encountered innatural 137 language, it also constitutes one of the oldest conundrums in temporal logic. The idea of historical 138 necessity denotes the intuition that, looking forward to the future from any given moment, it 139 seems that several different developments are possible. But once they happened, there is a sense 140 of necessity and finality to them—the future is open, but the past is settled once and forall.This 141 conundrum goes back at least to Aristotle and has been picked up by Prior (Prior, 1957, 1967), later 142 modeled by Thomason (1970, 1984) and used to model semantic phenomena such as the progressive, 143 e. g. Dowty (1977). These works on temporal logic rely on the ideaof branching time to derive the 144 asymmetry between the past and the future. In a branching-time model, times (or indices) do not 145 constitute a strict linear order. Instead, two indices that have a common predecessor may not be 146 ordered with respect to one another. The following definition is taken from [redacted]:

147 Definition 1 A branching-time frame U is a pair ⟨I, <⟩, where 148 1. I is a non-empty set of indices i; 149 2. < is an ordering on I such that if i1 < i and i2 < i, then either i1 = i2, or i1 < i2, or i2 < i1. 150 An index i1 is called a predecessor of i2 iff i1 < i2; it is a successor of i2 iff i2 < i1

151 All indices have a common predecessor.

152 A branch through any i ∈ I is a maximal linearly ordered subset of I containing i.

153 This definition creates a branching structure such that, given a specific index ic (the actual 154 present), there are several continuing developments branching off towards the future, but only 155 one line of indices that lead to ic. In other words, the future is open, but the past is settled. Tra- 156 ditional approaches to branching time are content with this binary distinction between the open 157 future and the settled past/present and limit quantification to those branches that pass through ic. 158 But [redacted] argues that the branching structure creates a three-way distinction, which natural 159 language expressions also refer to. In addition to the above definitions, we also assume that indices

4 ic

Figure 1: The three domains of the modal-temporal space, relative to the actual present ic: the actual (solid line); the possible (dashed lines); the counterfactual (dotted lines).

160 can be grouped into sets of simultaneous moments, so that they have a strictly ordered time value 161 in addition to their position within the tree-like frame. 162 The three modal domains we can formally distinguish are the following: 163 1. ic and predecessors of ic (the actual); 164 2. successors of ic (the possible); 165 3. and indices that are neither successors nor predecessors of nor identical with ic (the coun- 166 terfactual); 167 In this framework, we can define the domain referred to by irrealis expressions ex negativo, 168 as the set of indices that are not predecessors of or identical with the actual present. We can also 169 see that the domain of irrealis is composed of the possible (futures) and the counterfactual (past, 170 present and future). 171 The idea that branching time is instrumental in understanding the realis/irrealis distinction is 172 not new. Baker & Travis (1997) base their account on the assumption that speakers conceptualize 173 time as branching rather than linear, and so do McGregor & Wagner (2006). However, their ac- 174 counts follow the traditional approach to branching time in allowing only a two-way distinction 175 between the actual present and past as opposed to the possible futures, without affording exclusive 176 access to only counterfactual indices.² 177 On the other hand, several researchers working on languages with a realis/irrealis distinction 178 have recognized the need for a tripartite, rather than a binary, opposition, but have not considered 179 a branching-time framework for modeling this system. The initial quote by Sapir (1930) in the 180 previous section already refers to merely potential and the contrary to fact as two different types 181 of unreal activities. Another account that expands on this distinction is Roberts (1990), which is 182 summarized in table 1. His conclusions are based on an empirical study of the TAM systems of 183 Amele and other Papuan languages. 184 The same tripartite division is also acknowledge and confirmed in a later comparative studyof 185 realis/irrealis systems, published as Van Gijn & Gipper (2009). Similar intuitions about the relation 186 between time and mood have also been put forward elsewhere, for example in Van Linden & Ver- 187 straete (2008), who treat this link as a matter of pragmatic extension, however, rather than purely 188 semantic. ²In these accounts, counterfactual indices can in principle be accessed through shifting the perspective to the past. But then, they can no longer be differentiated from actual and possible indices (compare ?)

5 Unreal worlds(s)[sic] Real world (future) (non-future) true factually true in potentially not potentially true in real world but real world true in real true in unreal world (counterfactual) (positive) world false factually false in potentially not potentially false in real world but real world false in real false in unreal world (negative (negative) world counterfactual)

Table 1: The tripartite division into the real, the potential and the counterfactual in Roberts (1990: 398).

189 In sum, several researchers have previously acknowledged the potential of a branching-time 190 model for the analysis of the realis/irrealis distinction. Others have recognized the division of the 191 irrealis domain into possible futures and the counterfactual past, present and future. But so far, 192 no one has combined both insights to argue for modeling the subdivided irrealis domain by means 193 of a branching-time framework. We suggest that this theoretical innovation allows for greater 194 conceptual clarity in defining and analyzing TAM systems that are based on the realis/irrealis 195 distinction. 196 Before concluding this section, let us come back to the initial criticism leveled against irrealis 197 as a meaningful comparative concept. In part, this was based on the difficulty of giving it a succinct 198 definition. As Bybee (1998: 267) states:

199 A highly generalized notion such as “lacking in reality” is probably too abstract to be 200 of much communicative use.

201 Generally speaking, there is a widespread consensus that highly abstract notions, such as identi- 202 fiability by the speaker, can in fact be at the heart of grammaticalized categories in natural language, 203 such as definiteness. The definition of irrealis proposed in this section is still highly abstract,but 204 sufficiently precise to operationalize the categorization of individual clauses as belonging toone 205 of three temporal-modal categories, as we further discuss in section 3.3. 206 In the following sections, we will spell out the consequences of our assumptions for the criti- 207 cism leveled against the notion of irrealis as a typological category.

208 3 Binary and non-binary realis/irrealis systems

209 In this section, we address the argument that the realis/irrealis distinction is understood as a binary 210 opposition, but in many languages that have a marker labeled irrealis, this marker may not be in bi- 211 nary opposition to a realis marker. Bybee (1998) ascribes some of the currency of the realis/irrealis 212 distinction to the structuralist tendency to look for binary oppositions that are not matched by 213 observations on natural languages. Contradicting this claim, Michael (2014) argues that Nanti, an 214 Arawak language of Peru, does present a clear example of a language with a binary realis/irrealis 215 opposition, where each category behaves as expected. Cleary-Kemp (2014) states that Koro and 216 other languages of the Admiralties Islands, too, have a well-behaved binary distinction between 217 realis and irrealis mood.

6 Figure 2: Black dots: Languages with a binary realis/irrealis distinction (24); white dots: languages without a binary realis/irrealis distinction (46).

218 In this section we show that, within the Oceanic context, such binary systems are not at all 219 uncommon. Furthermore, those languages that have a more than binary distinction can largely 220 be divided into two cases: In the first case, they implement a more fine-grained modal-temporal 221 division of the irrealis domain. We will show that this situation is in fact expected according to our 222 theoretical assumptions. In the second case, certain functions typically associated with irrealis, 223 such as imperatives, are expressed by more specialized markers.

224 3.1 Binary realis/irrealis systems in Oceanic

225 As previously discussed, the Oceanic family of languages has been noted consistently to fea- 226 ture TAM systems that revolve around the realis/irrealis distinction. For an overview of how 227 widespread binary realis/irrealis systems are in this group of languages, we compiled a compara- 228 tive study on a convenience sample of 74 languages. We excluded four languages from the final 229 sample, because we could not determine with certainty from the literature whether they had a bi- 230 nary distinction. Of the remaining 70 languages, 24 have a binary system of realis/irrealis marking, 231 while 46 languages do not. The map in figure 2 suggests that binary systems are more widespread 232 within Melanesia compared to the rest of Oceania. None of the languages in our sample with a 233 binary system is spoken primarily outside of Melanesia. Moreover, within Vanuatu, there is a clear 234 concentration of binary systems in North and Central Vanuatu, especially around Malakula and 235 Santo islands, as shown in figure 3. This means that, within Oceanic languages, not all markthe 236 category of irrealis at all; and among those that do, not all leave it at a binary distinction between 237 realis and irrealis. At the same time, there is a substantial number of languages in the region that 238 do organize their TAM systems around a binary realis/irrealis distinction. 239 One example of a language with a binary realis/irrealis distinction is Tape, as described in 240 Crowley (2006: 139ff.). The template for finite verb phrases is given as follows:

241 (2) SUBJECT-MOOD MOOD-ASPECT NEGATIVE NUMBER INCEPTIVE ROOT OBJECT

242 This is illustrated by the following example:

243 (3) pe-ska-r-vin 1nonsg.incl:irr-neg-dl-go

7 Figure 3: Part of figure 2, zooming in on Vanuatu. Black dots: Languages with a binary realis/irrealis distinction; white dots: languages without a binary realis/irrealis distinction.

8 244 “we (dl. incl.) will not go” (Crowley, 2006: 139)

245 The prefixes in the first position of this verbal complex form a paradigm that simultaneously en- 246 codes person-numer distinctions in agreement with the subject and the realis/irrealis distinction. 247 Morphemes that encode two semantically and typologically different categories at the same time 248 are commonly referred to as portmanteau morphemes. The situation in Tape, where verbs are 249 prefixed by a portmanteau subject-mood affix is quite widespread within Oceanic and wewillsee 250 more such systems in section 4.3.

Realis Irrealis sg nonsg sg nonsg incl dë- incl ba- 1 en- 1 be- excl më- excl pa- 2 kë- ka- 2 po- këpa- 3 i- i- 3 ipo- ipa-

Table 2: The Tape system of portmanteau subject-mood prefixesCrowley ( , 2006: 142).

251 As can be seen from the verb complex schema in (2), both the mood value of the predicate 252 and the number of the subject can be further differentiated by subsequent affixes before the verb 253 root. In particular, predictions with a high degree of certainty, as well as obligations, are typically 254 expressed with a necessitative affix in the MOOD-ASPECT slot in (2). In this sense, Tape can be 255 said to have more than two markers with modal implications. However, the opposition between 256 realis and irrealis is implemented within the same morphological paradigm, while the necessita- 257 tive marker occurs in a different slot in the verbal complex. At the same time, a closer lookat 258 the examples in Tape shows that the “realis” set of subject-mood markers might be more accu- 259 rately characterized as being neutral with respect to modality, since they also occur, for example 260 in the apodosis of counterfactual clauses. We have evidence from similar cases in which one set of 261 subject-mood markers is simply neutral with respect to modality, in effect only encoding subject 262 agreement. We will return to those cases in section 4.3. For the purposes of our regional typo- 263 logical overview, however, we counted cases such as Tape as exhibiting a binary distinction, even 264 though this distinction may be one between neutral and irrealis rather than realis and irrealis. 265 Another example for a language we counted as having a binary realis/irrealis distinction is 266 Taiof, a language spoken on the eponymous island off Bougainville. The Taiof verbal complex has 267 three preverbal and one post-verbal slot for TAM marking:

268 (4) real/irr fut/ctf habit ROOT impf/prf

269 As with Tape before, we can see that Taiof can distinguish between more than just two modal 270 domains within its verbal complex. Yet, in the first slot of this structure, there is a binary opposition 271 between realis and irrealis markers. The description is too sparse to decide whether the values of 272 these markers are in fact best characterized as “realis” and “irrealis”. The occurrence of the realis 273 marker in the protasis of a counterfactual clause indicates that it might encode non-future temporal 274 reference instead. 275 While many languages in our sample have thus not been investigated in sufficient detail to 276 draw definitive conclusions, we do have sufficient insight into one language with a binary system, 277 which is Mavea, a language of Vanuatu. This language has been extensively described by Guérin

9 Singular Plural Dual Paucal/trial Realis Irrealis Realis/irrealis 1incl da- dar- datol- 1excl na- ka- ki- kir- kitol- 2 ko- ko- ki- kir- kitol- 3 mo- i- ra- rar- ratol-

Table 3: The paradigm of portmanteau subject agreement markers in Mavea, according to Guérin (2011: 211).

278 (2006b), and we also have access to a corpus comprising more than 64k tokens Guérin (2006a, 2019). 279 Within this corpus, several entire texts, comprising 639 clauses, were manually tagged for their 280 TAM values by two trained annotators. Mavea has a paradigm of portmanteau subject proclitics, 281 similar to Tape, except that the realis/irrealis distinction is only implemented for the first and third 282 person singular. 283 We searched the Mavea corpus for counterexamples to the generalization that realis subject- 284 mood proclitics only occur in realis contexts, and irrealis subject-mood proclitics only occur in 285 irrealis contexts, as defined above. We did not find such counterexamples and are therefore con- 286 fident that the two sets of portmanteau subject-TAM markers encode realis and irrealis reference 287 respectively. We also did additional targeted elicitations in Mavea, which further corroborated this 288 conclusion [redacted]. 289 Among those 46 languages without a binary realis/irrealis opposition are some languages that 290 do not have any obligatory TAM marking at all, some that primarily distinguish aspectual or tem- 291 poral rather than modal references within their TAM systems, and those that distinguish more 292 than one modal-temporal domains. We discuss some of these latter cases in the following sections.

293 3.2 Modal-temporal subdivisions in the irrealis domain

294 As we have argued in section 2, the domain of irrealis modality can further be divided into the 295 possible and the counterfactual. These subdivisions are the reason why some languages havea 296 non-binary contrast between realis and other modalities. An example is the Oceanic language 297 Daakaka. The verbal complex of this language is given in table 4.

sbj.agr (=)tma (aux)(redup-) Verb (-res) (=trans) na,… =m,… du,pwer … …… =ne

Table 4: Structure of the verbal complex in Daakaka (from von Prince et al. 2019).

298 The system of TAM markers in Daakaka is shown intable 5. The three main modal-temporal 299 distinctions in the system are between the (positive and negative) realis markers, the (positive and 300 negative) potential markers and the distal marker. The open polarity marker doo is restricted to 301 embedded polarity questions, and the change-of-state marker bwet is found only in realis contexts 302 with a corresponding aspectual denotation. In effect, while there is a realis marker, it is not opposed 303 to a single irrealis marker. Instead, the domain of irreality is split between the potential markers

10 ic

Figure 4: The main modal-temporal oppositions marked in Daakaka: realis (light-gray); potential (dark gray); distal (white with gray outline).

304 and the distal marker, is the main topic of this section. In addition, there is interaction with polarity 305 and with aspect, which we will also consider briefly in section 4.3.

enclitic proclitic monosyllabic Pos. Realis =m mw= mwe/mV Neg. Realis to Pos. Potential =p w= wV Negative Potential =n n= nV Distal =t t= tV Open Polarity doo Change of State bwet

Table 5: The system of TAM markers, adapted from von Prince (2015). V: vowel; Pos: positive; Neg: negative.

306 Through a targeted elicitation study in Daakaka, we have shown in [redacted] that the division 307 of labor between the realis, the potential and the distal marker in Daakaka is roughly as follows: 308 The distal marker overlaps with the realis in that it also refers to the actual past, in additionto 309 the counterfactual past and present. The potential marker is restricted to future contexts and the 310 possible present. This picture is summarized in figure 4. 311 Thus, the distal marker is found, for example in contexts expressing possibilities of thepast,or 312 the counterfactual present, as shown in (5) and (6).

313 (5) ma wese ka te me yan vilye s-an vi real enough comp dist come on place cl3-al.pl white.man 314 “it may have come from a Western country” (lit. “…from the place of white men”) (4104)

315 (6) na=t ka pini or. 1sg=dist fly fill place 316 “[I wish I had wings,] I would fly around everywhere.” (4209)

11 317 Future possibilities, by contrast are expressed by the potential marker:

318 (7) ko=m kuowilye ka ko=p vinye ne vis ane tes 2sg=real know comp 2sg=pot shoot with weapon trans sea 319 “You can shoot [sharks] with a harpoon” (1741)

320 Daakaka therefore represents languages which do not implement a purely binary division be- 321 tween realis and irrealis domains, but which carve up the irrealis domain into smaller subdomains. 322 The three-way distinction between realis, potential and distal corresponds roughly, butnotcom- 323 pletely to the three-way distinction between the actual, the counterfactual and the possible (com- 324 pare [redacted]).

325 3.3 Functional subdivisions of the irrealis domain

326 The modal-temporal domains outlined in section 2 are not the only dimensions along which the 327 irrealis domain can be further divided. Irrealis markers can also be described in terms of their 328 discourse functions and the sentence types they occur in. Thus, the temporal-modal domain of 329 possible futures is associated with various clause types, ranging from assertions about the distant 330 and imminent future, to imperatives, prohibitives, purpose clauses and complements of verbs like 331 want. Some languages will use an irrealis or relative-future marker to cover all these functions. 332 But others may have more specialized markers for individual functions. In Oceanic languages, this 333 applies particularly often to the following functions: 334 1. imperatives; 335 2. prohibitives; 336 3. timitives; 337 4. immediate futures. 338 Imperatives and prohibitives are specific clause types associated with directive speech acts. In 339 a sample of 54 Oceanic languages 26 use an irrealis marker. In 18 languages, both positive and 340 negative directives are expressed by only a bare verb.³ This set includes languages such as Banomi, 341 in which bare verbs are also the default form in other irrealis environments. Nine languages in our 342 sample use one TAM marker that is specific for imperatives, or one for prohibitives, or two markers, 343 one for each type of directive. We will discuss two examples showing the latter case below. Ifira- 344 Mele was described as using an intentional marker for imperatives (Clark, 2002: 688), although the 345 existing accounts are too scarce for a precise characterization of this marker. Rotuman not only 346 has a special marker for prohibitives, but also optionally uses word-order inversion from SV to VS 347 to mark imperatives. These results are summed up in table 6. 348 Timitive modality is sometimes also referred to as apprehensive or aversive modality. Dedi- 349 cated markers for this category are relatively widespread among Oceanic languages (Lichtenberk, 350 2016). It is characterized by a combination of epistemic possibility and undesirability (they might 351 fight). Timitives partially overlap with prohibitives in their function. A warning to the addressee 352 about some possible, undesirable event beyond their immediate control, such as (watch out), don’t 353 fall, is simultaneously prohibitive and timitive. In some languages, this relation between timitives 354 and prohibitives corresponds to identical or similar markers for both functions (cf. Smith-Dennis, 355 2019). 356 Immediate futures are used to refer to imminent events and actions. In languages that distin- 357 guish between imminent and non-imminent futures, it is not always clear from the description

³Languages that use an irrealis marker to express directives generally also allow bare verb forms as imperatives. We still counted those as using irrealis.

12 TAM # irrealis 26 bare 18 special 9 intentional 1 Total 54

Table 6: Strategies of expressing imperatives and prohibitives in a convenience sample of Oceanic languages.

358 whether this contrast instantiates a graded-tense system, or whether the imminent future marker 359 would be more aptly characterized as an aspectual category (compare Copley, 2002, 2009). For our 360 current purposes, the distinction is not crucial. 361 Table 7 shows how different functions map to temporal-modal sub-domains of irrealis. Func- 362 tions above the double rule are typically encoded by a single form in Oceanic languages, while 363 functions underneath it are often expressed by more specialized markers. 364 Note that future markers in Oceanic languages typically express relative future, that is they are 365 relative to the topic time, not necessarily the utterance time (compare Klein, 1994). Thus in sen- 366 tences such as she wanted/was able/had [to run], the embedded predicate will typically be marked 367 as future/irrealis even when the event time it refers to is prior to the time of utterance. In the 368 literature, such markers are variably referred to as relative futures or as prospective aspect. We 369 choose the former term, but this choice does not come with a strong committment; our approach 370 is compatible with either analysis. 371 One example for a TAM system that distributes the range of irrealis functions over several 372 different markers is Paamese. Crowley (1982) characterizes the system as shown in table 8. 373 In addition to a zero-marked realis category, Paamese has one form to refer to the immedi- 374 ate future, one for the distant future, one for prohibitives and one for imperatives. According to 375 Crowley (1982: 133), the “potential” form “indicates that a non-real event may become real. It is 376 often used as a warning that something the addressee regards as unpleasant might happen…”. An 377 example is given in (8):

378 (8) Sim nemavul 379 sii+mo na+mavulu bone.2sg 3sg.pot.break

380 “Your bones might might break” (i.e. “Watch out or your bones might break”)

381 The description and the example suggest that the Paamese potential marker primarily expresses 382 timitive modality. The closely related language Vatlongos (Southeast Ambrym), spoken onthe 383 neighbouring island of Ambrym, has a very similar system. Ridge (2019) describes it as follows: 384 • Realis: 385 – Prior 386 – Non-future 387 • Irrealis: 388 – Immediate Future 389 – Distant Future

13 Function Example Modal-temporal domain Fut. assertions It will rain tomorrow. Possible fut. If you had been quiet, they would not have CNTR past found us. Conditional If you were quiet, they would not find us. CNTR pres./fut. If you are quiet, they won’t find us. Possible pres./fut. If you were quiet, they didn’t see us. Possible past want I want [to visit Beijing]. Possible (rel.) future wish I wish [I had tried harder/had more CNTR past/pres./fut. money/could visit Beijing] Ability She can make pasta. Possible (rel.) fut. Obligation We have to leave. Possible (rel.) fut. Purpose She shouted [to get their attention]. Possible (rel.) fut. She ran [lest she miss the train]. Possible (rel.) fut. Timitive Watch out [you might fall]. Possible (rel.) fut. They might be/have been sick. Possible pres./past Imperative Sit down! Possible fut. Prohibitive Don’t touch that! Possible fut. Imm. future I’m leaving. Possible fut.

Table 7: Mappings from functions to modal-temporal subdomains of irrealis; inspired by Bugenhagen (1993).

390 – Apprehensive 391 – Imperative 392 – Prohibitive 393 Cases such as these show that the distinction between realis and irrealis modalities can play a 394 role in a language even though the irrealis domain (and to a lesser extent, the realis domain) may 395 be subdivided according to specific functions and clause types. 396 Before concluding this section, we revisit two more contexts that have been associated with the 397 domain of irrealis, mainly through Chafe (1995) and Mithun (1995), namely negation and questions. 398 In Oceanic, we do find a mild interaction between mood and negation. However, in the Oceanic 399 context, it is not the case that realis mood is incompatible with negation – in contrast to Caddo 400 and Nanti, as described in Chafe (1995) and Michael (2014) respectively. Instead, there are typically 401 two sets of negative markers, one for realis, one for irrealis mood. Negation should interact with 402 the marking of reality status as the proposition that is negated is claimed to be not real, hence 403 if negation scopes over mood we should expect irrealis forms. However, if mood scopes over 404 negation, we might find either realis or irrealis forms. It seems that we find the latter situationin 405 Oceanic languages, with a twist: markers that combine reality status and negation. Thus, we have 406 seen in section 3.2 that Daakaka has a positive and a negative set of realis and potential markers. 407 In a convenience set of 51 Oceanic languages, we found four others with a similar pattern. These

14 Crowley (1982: 129) Notes realis realis (zero-marked) immediate immediate future distant distant future potential timitive prohibitive prohibitive imperative imperative

Table 8: TAM categories in Paamese, according to Crowley (1982), with notes about cor- respondences to terms used in this article.

Ulithian Puluwatese

Rotuman Merei Daakaka

Figure 5: Black dots: Languages which show significant interaction between negation and mood marking. White dots: Languages from our sample that do not show such interaction.

408 are highlighted in figure 5. 409 Questions, on the other hand are neutral with respect to the realis/irrealis distinction inOceanic. 410 The only exception to this rule that we found in a survey of 67 languages are embedded polarity 411 questions in Daakaka, for which there is a special TAM marker (compare section 3.2). By and large, 412 irrealis is not associated with negative utterances and with questions in Oceanic.

413 3.4 Summary

414 We have seen that in many Oceanic languages, the TAM system does in fact center around a binary 415 realis/irrealis divide. For those that have a realis/irrealis distinction in a non-binary system, we 416 have shown that the irrealis category is often split up into more fine-grained domains. In sum, the 417 systems we see are fully compatible with the idea that the realis/irrealis distinction can be central 418 in the organization of TAM systems, even in languages that further subdivide the irrealis domain 419 along its temporal-modal dimensions or discourse functions.

15 420 4 Cross-linguistic mismatches

421 4.1 Overview

422 The second main argument against the notion of irrealis as a meaningful category in cross-linguistic 423 comparison states that the label irrealis is sometimes assigned to elements of different languages 424 that do not have any functional overlap. In this section, we argue that the causes for this are 425 largely the same as laid out in sections 3.2 and 3.3. The phenomenon is illustrated quite succinctly 426 by a pair of languages discussed by de Haan (2012), Limbu (Tibetu-Burman) and Hualapai (Pai, 427 Yuman). Both languages have a marker labelled as irrealis. But in Limbu, this marker appears to 428 be restricted to counterfactual situations of the present or past, while the Hualapai irrealis only 429 refers to the immediate future. This contrast is shown by the following two examples from de 430 Haan (2012: 108):

431 (9) Limbu (Tibeto-Burman) 432 yaŋ kɔtt-u-ŋ-gɔ:ni iŋ-u-ŋ-ba. money have-3p-1sg.ag-irr buy-3p-1sg.ag-ipf 433 “If only I had the money, I would buy it.”

434 (10) Hualapai (Pai, Yuman) 435 olo-h-ch ha: thi:-hi-k-wi horse-dem-sub water 3/3.drink-irr-ss-aux/be 436 “The horse is going to drink the water.”

437 In effect, the Limbu “irrealis” marker has no overlap at all with the Hualapai “irrealis” marker. 438 In this particular instance, the approach we have developed so far is sufficient to account for the 439 mismatch between labels: We have already seen that the domain of irrealis can be divided into 440 more fine-grained temporal and modal subdomains. Markers of individual languages might refer 441 to only one of those subdomains. Thus, the two “irrealis” markers can be understood as referring 442 to two separate, non-overlapping subdomains of irrealis, as illustrated in figure 6.

ic ic

Figure 6: Suggestions for the references of the Limbu irrealis (left) and the Hualapai irrealis (right).

443 In cases like this, it is clear that the label “irrealis” is not entirely inappropriate for the respective 444 markers, but ultimately misleading. For purposes of cross-linguistic comparison, it would be better 445 to use more precise terms that correspond more closely to the semantics of each marker, such as 446 counterfactual (past/present) in the case of Limbu, or immediate future in the case of Hualapai. 447 Many, if not most cross-linguistic mismatches in how the label irrealis is applied can be traced 448 back to this type of situation. There are two more scenarios that frequently lead to inconsistencies 449 in how realis and irrealis markers behave across languages: 450 1. Irrealis/realis markers missing expected meanings: A general realis or irrealis marker does

16 451 not display the expected range of meanings, because some meanings are blocked by a more 452 specific marker of the same paradigm. 453 2. Irrealis/realis markers occurring outside of their respective domain: In a system with both 454 realis and irrealis forms, one of the two might in fact be neutral with respect to TAM. This 455 can explain why, for example, a marker labeled as “realis” may occur in irrealis environments 456 such as counterfactual conditionals. 457 Both of these scenarios are further explored in the following sections.

458 4.2 The effects of blocking

459 We have seen in sections 3.2 and 4.1 that temporal-modal markers may refer to only a subdomain 460 of the irrealis domain. We have analyzed the corresponding cases as involving lexically determined 461 differences between markers. In some other cases, however, a more plausible approach is toassume 462 pragmatically conditioned restrictions to the interpretation of irrealis markers. When a marker A 463 covers the entire range of irrealis meanings except one, and there is another marker B in the 464 language that exclusively expresses this specific meaning, marker A may in fact be an irrealis 465 marker, whose range of interpretations is restricted by blocking. 466 The notion of blocking has been used in various contexts to explain the behaviour of certain 467 elements within the same morpho-syntactic paradigm in terms of their paradigm-internal compe- 468 tition.⁴ 469 To illustrate with an example by de Saussure (Third course of lectures on General Linguistics 470 (1910-11), Chung 1994): In a language that has a singular, dual and plural form, the plural can be 471 seen as applying to two or more entities, but its use to refer to two entities is blocked because the 472 dual, as the more specific form, is used (also compare Kiparsky, 2005). 473 Nakanai has been described as having two irrealis markers, ge and ga (Johnston, 1980: 63ff.). 474 The “non-imminent irrealis” marker ge covers a wider range of contexts, including future refer- 475 ence, hypothetical and counterfactual conditionals, complements of “want”, purpose clauses and 476 expressions of possibility. This is illustrated in (11):

477 (11) Eia ge tuga. 3sg nim.irr depart 478 “He will/might/could/should depart.” (Johnston 1980: 63, glossed by authors)

479 The “imminent irrealis” marker ga, by contrast, is reserved for contexts referring to imminent 480 or frustrated actions:

481 (12) a. Eau ga la-lea. 1sg im.irr redup-sick 482 “I’m getting sick.”

⁴It is this sense of the term that we are appealing to here. There have also been other uses of this term. Thus, in Smolensky & Prince (1993) and much of the early literature on Optimality Theory, it denotes a relation between rules rather than a relation between expressions. In generative syntax, blocking can denote the property of a syntactic element that prevents the movement of another element.

17 ic

Figure 7: The two irrealis markers of Nakanai: the “non-imminent irrealis” (shaded dark gray) and the “imminent irrealis” (shaded white).

483 b. Eau ga tuga so-io, ouka . 1sg im.irr walk to-there no 484 “I was about to proceed on, but didn’t (i. e. because some thing prevented me).” (John- 485 ston 1980: 64, glosses adjusted from irr to im.irr)

486 We can model the difference between the two irrealis markers in Nakanai as indicated infigure 487 7: The imminent irrealis marker refers to the indices directly preceding (and possibly including) 488 ic, where ic is the topic time, rather than the utterance time. The non-imminent irrealis marker 489 appears to cover the entire irrealis domain, but is not used in contexts of imminent future refer- 490 ence. We assume that, in these contexts, its use is blocked by the availability of the more specific 491 imminent marker. 492 We would like to note here that the Nakanai “imminent irrealis” marker might be amenable to 493 an analysis similar to what Copley (2009) proposes for English going to, which involves progressive 494 aspect in addition to a future reference. Our treatment here is not meant as an exhaustive seman- 495 tic analysis of the Nakanai markers, but rather to serve as an illustration how one more specific 496 expression might restrict the range of an irrealis marker through blocking. 497 The TAM system of Sisiqa (Solomon islands) appears remarkably similar to Nakanai intermsof 498 dividing the irrealis domain into realis, irrealis and imminent future. It has a system of 33 subject- 499 TAM portmanteau markers that feature four person distinctions and three number distinctions in 500 addition to the tripartite TAM values (Ross, 2002: 462). 501 In short, some irrealis markers do not show the full range of functions associated with irrealis. 502 In cases like the Nakanai non-imminent irrealis, this function is a reference to the immediate fu- 503 ture. Its absence can be derived from the fact that Nakanai has a more specialized marker that is re- 504 stricted to this particular function and therefore blocks the use of the more general (non-imminent) 505 irrealis marker.

506 4.3 Realis, irrealis and unspecified

507 Finally, markers labeled as realis sometimes show a wider distribution than expected. We argue 508 that, in these cases, the label may have been applied to forms that are in fact neutral with respect 509 to TAM values. Their usual interpretation as expressing realis reference derives from pragmatic 510 defaults and from their contrast to other forms in the same system. 511 For example, Nafsan (Vanuatu) has been analyzed as expressing realis and irrealis by portman-

18 512 teau subject proclitics (Thieberger, 2006), similar to what we have seen in section 3.1. However, in 513 her work on Nafsan, [author2] found that the subject proclitics labeled as realis appear in many 514 contexts that should be incompatible with realis meanings [redacted]. A good illustration of con- 515 texts with irrealis meanings in which realis subject proclitics are found are the combinations of 516 subject proclitics and TAM markers in two Nafsan corpora (Thieberger 2006, [redacted]). While 517 the irrealis and perfect subject proclitics occur only with the markers that match their TAM values, 518 the “realis” proclitics can occur with almost all TAM markers. For instance, the realis proclitics can 519 combine with the perfect pe in all perfect contexts, including those with future reference, as shown 520 in (13).

521 (13) Malnen p̃a=ler, a=pe mtir natus su. when 2sg.irr=return 1sg.real=prf write letter prf 522 “When you come back, I will have finished writing the letter.” ([redacted], based on Dahl 523 2000: FQ 17)

524 There are also other cases of the realis proclitics being used with future reference, asin (14) 525 and (15):⁵

526 (14) Ko ru=fla, ru=fla, mees i=fla ta nom, i=fla or 3pl.real=might 3pl.real=might today 3sg.real=might neg1 finish 3sg.real=might 527 ta nom mau, matol ru=mai, pkaskei pa pnut nawesien gar. neg1 finish neg2 tomorrow 3pl.real=come same go close work 3pl.ben 528 “And if it is not finished today, if it is not all done, tomorrow they’ll come to their work.” 529 (085.017)

530 (15) [There are black clouds in the sky.] It RAIN (very soon).(Dahl, 2000: FQ 47) 531 Us i=wo pelpel. rain 3sg.real=rain soon 532 “It will rain soon.” (AK1-086-01)

533 Another striking context in which realis proclitics are very frequent are protases of counter- 534 factual conditional clauses with the conditional marker f, as in (16). While in counterfactual con- 535 ditional protases both realis and irrealis subject proclitics are felicitous, only irrealis is felicitous in 536 the apodosis of (16).

537 (16) a=f mer mes matol, go nfag nen kin a=tai nakn-i-k 1sg.real=cond cf play tomorrow and sore rel comp 1sg.real=cut finger-v-1sg.poss 538 ke=fo mer makot 3sg.irr=psp.irr again break 539 “If I played tomorrow, the sore I cut on my finger would bleed again.” (AK1-098-01, 540 00:03:39.185-00:03:57.063)

541 The fact that “realis” proclitics appear in irrealis contexts in (13)-(16) shows that they are not re- 542 stricted to the past and present reference of the actual world and should not be analyzed as encod- 543 ing the realis mood. A similar situation is attested in Wogeo, a Western Oceanic language spoken 544 on the island of Wogeo in the North of Papua New Guinea. Just like Nafsan, Wogeo has realis

⁵The choice of the general proclitic in (15) might be related to the choice of the temporal adverb. Ex- amples including “in a few minutes”, “in the evening”, and “tomorrow” were produced with irrealis and the prospective irrealis fo.

19 545 and irrealis subject proclitics (Exter, 2012), outlined in Table 9. Exter (2012) reports on differences 546 between realis and irrealis subject markers: Irrealis is chosen in non-actual environments, such as 547 future and different kinds of possibilities (17) and realis is chosen for the past and present reference 548 (18).

Table 9: Subject proclitics in Wogeo from Exter (2012) Realis Irrealis 1sg o- go- 2sg go-, ko- go- 3sg e- de- 1du to- tog- 2du kad- 3du do- dog- 1pc to- tog- 2pc koto- 3pc doto- 1pl ta- 2pl ka- 3pl da-

549 (17) go-lako 1sg.irr-go 550 “I must go”, “I want to go”, “I will go (now)” (Exter, 2012: 182)

551 (18) o-lako 1sg.real-go 552 “I go”, “I went” (Exter, 2012: 182)

553 The problem with the realis/irrealis definition of Wogeo subject markers is that realis andir- 554 realis are completely interchangeable in occurring with the future (19) and tentative markers (20), 555 without a change in meaning. Moreover, realis appears obligatorily in protases of counterfactual 556 conditionals (21) because irrealis is not morphologically available with the counterfactual marker 557 s- (Exter, 2012: 186). We can see, however, that realis subject markers are also optionally used in 558 the apodosis of the conditional in (21). Exter (2012: 184) mentions several other modal contexts in 559 which realis markers occur, such as ability, permission, and protases and apodoses of hypothetical 560 conditionals. The only meanings restricted to irrealis are obligation, volition, and immediate future 561 (Exter, 2012: 185).

562 (19) a. m-o-lako fut-1sg.real-go 563 b. mo-go-lako fut-1sg.irr-go 564 “I will/can/may go.” (Exter, 2012: 182)

20 565 (20) a. s-o-lako tent-1sg.real-go 566 b. so-go-lako tent-1sg.irr-go 567 “I try it by going.” (Exter, 2012: 182)

568 (21) s-e-vá iko sa-k-lako, katé cf-3sg.real-happen you cf-2sg.real-go thus 569 mo-la-moet Glossfut.Gloss2sg.Glossreal-inch-disappear 570 “If you had gone, you would have been lost.” (Exter, 2012: 186)

571 The fact that subject markers that have been described as encoding realis can appear with 572 numerous future and modal meanings in Nafsan and Wogeo, including counterfactuals, tells us that 573 the “realis” paradigms in these languages should better be analysed as a subject marking of person 574 and number only, semantically underspecified for TAM values. In both languages this reanalysis 575 can explain why in some contexts the two paradigms of subject markers are interchangeable or 576 why only the general marking is available due to morphological constraints. 577 The reason these unspecified subject markers often receive realis past and present meanings lies 578 in pragmatic defaults of temporal reference. According to Smith et al. (2007), the utterance time 579 is the central orientation point for language, which is pragmatically the simplest kind of temporal 580 reference (Mucha, 2015). Since past and present temporal references are (epistemically) closer to 581 the utterance time than the future, they are the pragmatically default temporal interpretations of 582 verbs unspecified for TAM. This reanalysis of realis subject proclitics confirms some ofthedoubts 583 raised by Cristofaro (2012), who advocates caution when diagnosing subject markers as expressing 584 the realis/irrealis distinction.

585 5 Conclusion

586 We have discussed the validity of the notion of irrealis as a semantic concept and a cross-linguistic 587 category. We find that, at least in the context of Oceanic languages, this notion is both usefulin 588 characterizing TAM systems, and meaningful in terms of understanding the underlying functions 589 of these systems. We have reviewed the main points of criticism that have been levelled against 590 this notion, most prominently by Bybee et al. (1994) and Bybee (1998), concerning its conceptual 591 appeal, the alleged prevalence of irrealis in non-binary mood systems, and cross-linguistic mis- 592 matches in how the label is applied. We have put forward a formal definition of irrealis in terms 593 of a tripartite branching-time frame. By looking at the comparative picture within Oceanic, we 594 showed that a binary realis/irrealis distinction is by no means exceptional, and that the irrealis is 595 not closely associated with either negation or questions. Through in-depth studies of individual 596 languages, we further showed how modal-temporal and functional subdivisions within the irrealis 597 domain account both for non-binary TAM systems which are still centred around the realis/irrealis 598 distinction; and for cross-linguistic mismatches in using irrealis as a label. We identified paradig- 599 matic competition, or blocking, and mood-neutral subject-proclitics in paradigms with portman- 600 teau irrealis subject proclitics as further reasons for cross-linguistic mismatches. Our theoretical 601 work provides for semantic distinctions which may facilitate more precise applications of this label 602 in future language descriptions. We conclude that irrealis is in fact quite a wide and abstract notion, 603 which none-the-less allows for meaningful description of individual languages and cross-linguistic 604 comparison.

21 605 References

606 Baker, Mark, & Travis, Lisa. 1997. Mood as verbal definiteness in a “tenseless” language. Natural 607 Language Semantics, 5, 213–269. 608 Barbour, Julie. 2011. Exploring mood in Neverver. Oceanic Linguistics, 50(1), 198–220. 609 Bhat, D. N. S. 1999. The prominence of tense, aspects, and mood. John Benjamins. 610 Bickerton, Derek. 1975. Dynamics of a creole system. Cambridge University Press Cambridge. 611 Blewett, S.C. 1991. Irrealis in Manam discourse. Language and Linguistics in Melanesia, 22(1-2), 612 1–20. 613 Bugenhagen, Robert D. 1993. The semantics of irrealis in the Austronesian languages of Papua 614 New Guinea. Pages 1–39 of: Reesink, Ger P. (ed), Topics in descriptive Austronesian linguistics. 615 Leiden: Rijksuniversiteit Leiden. 616 Bybee, J. L, Perkins, Revere, & Pagliuca, W. 1994. The evolution of grammar: Tense, aspect, and 617 modality in the languages of the world. The University of Chicago Press. 618 Bybee, Joan L. 1998. “Irrealis” as a grammatical category. Anthropological Linguistics, 40(2), 257– 619 271. 620 Bybee, Joan L., & Fleischman, Suzanne. 1995. Modality in grammar and discourse. Typological 621 studies in language. John Benjamins. 622 Callaghan, Catherine A. 1998. Lake Miwok irrealis. Anthropological Linguistics, 40(2), 228–233. 623 Capell, Arthur. 1971. The Austronesian languages of Australian New Guinea. Current trends in 624 linguistics, 8, 240–340. 625 Chafe, Wallace. 1995. The realis-irrealis distinction in Caddo, the Northern Iroquoian languages 626 and English. In: Bybee & Fleischman (1995). 627 Chung, Karen Steffen. 1994. Saussure’s third course of lectures on general linguistics (1910-1911), 628 from the notebooks of emile constantin/trosième cours de linguistique générale (1910-1911) 629 d’après les cahiers d’emile constantin. ed. and transl. by eisuke komatsu & roy harris. Histo- 630 riographia linguistica, 21(3), 448–451. 631 Chung, Sandra, & Timberlake, Alan. 1985. Tense, aspect, and mood. language typology and syntactic 632 description: Grammatical categories and the lexicon, vol. by timothy shopen, 202-258. 633 Clark, Ross. 2002. Ifira-mele. Pages 681–693 of: Lynch, John und Ross, Malcolm und Crowley Terry 634 (ed), The oceanic languages. Curzon Language Family Series. Richmond, Surrey: Curzon. 635 Cleary-Kemp, Jessica. 2014. Irrealis as verbal non-specificity in Koro (Oceanic). Annual Meeting of 636 the Berkeley Linguistics Society, February, 20–41. 637 Comrie, Bernard. 1985. Tense. Cambridge Textbooks in Linguistics. Cambridge University Press. 638 Condoravdi, Cleo. 2002. Temporal interpretation of modals: Modals for the present and for the 639 past. Pages 59–88 of: Beaver, David, Casillas, L., Clark, Brady, & Kaufmann, Stefan (eds), The 640 construction of meaning. Stanford, CA: CSLI Publications. 641 Copley, Bridget. 2009. The semantics of the future. New York, London: Routledge. 642 Copley, Bridget Lynn. 2002. The Semantics of the Future. Doctorate, Massachusetts Institute of 643 Technology. 644 Cristofaro, Sonia. 2012. Descriptive notions vs. grammatical categories: Unrealized states of affairs 645 and ‘irrealis’. Language sciences, 34(2), 131 – 146. Papers selected from the ’What do languages 646 encode when they encode reality status?’ workshop at the 41st Annual Meeting of the Societas 647 Linguistica Europaea, Forlì, Italy, 17-20 September 2008. 648 Crowley, Terry. 1982. The Paamese language of Vanuatu. The Australian National University: Pacific 649 Linguistics. 650 Crowley, Terry. 2006. Tape: a declining language of Malakula (Vanuatu). Pacific Linguistics, Re- 651 search School of Pacific and Asian Studies, The Australian National University. Edited byJohn

22 652 Lynch. 653 Dahl, Östen. 2000. The future questionnaire. Pages 789–799 of: Dahl, Östen (ed), Tense and aspect 654 in the languages of Europe. Berlin: Mouton de Gruyter. 655 Dempwolff, Otto. 1939. Grammatik der Jabem Sprache auf Neuguinea. Hamburg: Friederichsen de 656 Gruyter. 657 Dempwolff, Otto, Bradshaw, Joel, & Czobor, Francisc. 2005. Grammar of the Jabêm language in 658 New Guinea. University of Hawai’i Press. https://www.jstor.org/stable/20006782. 659 Dixon, Robert MW. 1980. The languages of australia. Cambridge: Cambridge University Press. 660 Dowty, David. 1977. Toward a semantic analysis of verb aspect and the English ‘imperfective’ 661 progressive. Linguistics and Philosophy, 1, 45–77. 662 Elliott, Jennifer. 2000. Realis and irrealis: Forms and concepts of the grammaticalisation ofreality. 663 Linguistic Typology, 4, 55–90. 664 Exter, Mats. 2012. ’realis’ and ’irrealis’ in wogeo: A valid category. Language Documentation & 665 Conservation Special Publication, 5, 174–190. 666 Foley, William A. 1986. The Papuan languages of New Guinea. Cambridge University Press. 667 Givón, Talmy. 1982. Tense-aspect-modality: The creole prototype and beyond. Pages 115–163 668 of: Hopper, Paul J. (ed), Tense-aspect: Between semantics and pragmatics. Amsterdam: John 669 Benjamins. 670 Givón, Talmy. 1994. Irrealis and the subjunctive. Studies in Language, 18, 265–337. 671 Guérin, Valérie. 2006a. Documentation of Mavea. London: SOAS, Endangered Languages Archive. 672 https://elar.soas.ac.uk/Collection/MPI67426. [Accessed on 2017/03/01]. 673 Guérin, Valérie. 2006b. Mavea (Vanuatu). recordings and texts. Digital collection managed by ELAR. 674 ELDP Id: IGS0031. 675 Guérin, Valérie. 2011. A grammar of Mav̋ea: An Oceanic language of Vanuatu. Honolulu: University 676 of Hawai’i Press. 677 Guérin, Valérie. 2019. Storyboards in mavea. to be archived. 678 de Haan, Ferdinand. 2012. Irrealis: fact or fiction? Language Sciences, 34(2), 107 – 130. Papers 679 selected from the ’What do languages encode when they encode reality status?’ workshop at 680 the 41st Annual Meeting of the Societas Linguistica Europaea, Forlì, Italy, 17-20 September 2008. 681 Iatridou, Sabine. 2000. The grammatical ingredients of counterfactuality. Linguistic Inquiry, 31(2), 682 231–270. 683 Johnston, Raymond Leslie. 1980. Nakanai of New Britain: The grammar of an Oceanic language. 684 The Australian National University: Pacific Linguistics. 685 Kinkade, Dale. 1998. Is irrealis a grammatical category in Upper Chehalis? Anthropological Lin- 686 guistics, 40(2), 234–244. 687 Kiparsky, Paul. 2005. Blocking and periphrasis in inflectional paradigms. Dordrecht: Springer 688 Netherlands. Pages 113–135. 689 Klein, Wolfgang. 1994. Time in language. London, Berlin: Routledge. 690 Lichtenberk, Frantisek. 2016. Modality and mood in Oceanic. Chap. 14, pages 330–361 of: Nuyts, 691 Jan, & van der Auwera, Johan (eds), The Oxford Handbook of Mood and Modality. Oxford: Oxford 692 University Press. 693 Van Linden, An, & Verstraete, Jean-Christophe. 2008. The nature and origins of counterfactuality 694 in simple clauses: Cross-linguistic evidence. Journal of Pragmatics, 40, 1865–1895. 695 Martin, Laura. 1998. Irrealis constructions in Mocho (Mayan). Anthropological Linguistics, 40(2), 696 198–213. 697 McGregor, William B., & Wagner, Tamsin. 2006. The semantics and pragmatics of irrealis moodin 698 nyulnyulan languages. Oceanic Linguistics, 45(2), 339–379. 699 Michael, Lev. 2014. The Nanti reality status system: Implications for the typological valitity ofthe

23 700 realis/irrealis contrast. Linguistic Typology, 18(2), 251–288. 701 Mithun, Marianne. 1995. On the relativity of irreality. In: Modality in grammar and discourse. John 702 Benjamins. 703 Mucha, Anne. 2015. Temporal interpretation and cross-linguistic variation. Ph.D. thesis, Unversität 704 Potsdam. 705 Palmer, F.R. 1986. Mood and modality. 2 edn. Cambridge: Cambridge University Press. 706 von Prince, Kilu. 2015. A grammar of Daakaka. Berlin, Boston: De Gruyter Mouton. 707 von Prince, Kilu, Krajinović, Ana, Margetts, Anna, Thieberger, Nick, & Guérin, Valérie. 2019. Ha- 708 bituals in four Oceanic languages of Melanesia. Stuf - language typology and universals, 72(1), 709 21–66. 710 Prior, Arthur Norman. 1957. Time and modality. Oxford University Press. 711 Prior, Arthur Norman. 1967. Past, present and future. Oxford University Press. 712 Ridge, Eleanor. 2019. Variation in Vatlongos verbal morphosyntax: speaker communities in Southeast 713 Ambrym and Mele Maat. Ph.D. thesis, SOAS, London. 714 Roberts, John R. 1990. Modality in Amele and other Papuan languages. Journal of Linguistics, 16, 715 363–401. 716 Ross, Malcolm. 2002. Sisiqa. Pages 456–466 of: Lynch, John und Ross, Malcolm und Crowley Terry 717 (ed), The Oceanic languages. Curzon Language Family Series. Richmond, Surrey: Curzon. 718 Sapir, Edward. 1930. Southern Paiute, a Shoshonean language. Pages 1–296 of: Proceedings of the 719 American Academy of Arts and Sciences, vol. 65. JSTOR. 720 Smith, Carlota, Perkins, Ellavina, & Fernald, Theodore. 2007. Time in Navajo: Direct and indirect 721 interpretation. International Journal of American Linguistics, 73(1), 40–71. 722 Smith-Dennis, Ellen. 2019 (June). Don’t be fearful, lest it be undesirable. prohibitives and precautions 723 in Papapana. In: Apll 11. 724 Smolensky, Paul, & Prince, A. 1993. Optimality theory: Constraint interaction in generative gram- 725 mar. Optimality theory in phonology, 3. 726 Thieberger, Nick. 2006. A grammar of South Efate: an Oceanic language of Vanuatu. Oceanic 727 Linguistics Special Publications, no. 33. Honolulu: University of Hawai’i Press. 728 Thomason, Richmond H. 1970. Indeterminist time and truth-value gaps. Theoria, 36(3), 264–281. 729 Thomason, Richmond H. 1984. Combinations of tense and modality. Handbook of philosophical 730 logic, 2, 135–165. 731 Trask, Robert Lawrence. 1993. A dictionary of grammatical terms in linguistics. London, New York: 732 Routledge. 733 Van Gijn, Rik, & Gipper, Sonja. 2009. Irrealis in yurakaré and other languages. Cross-linguistic 734 semantics of tense, aspect and modality, 148, 155. 735 Verstraete, Jean-Christophe. 2005. The semantics and pragmatics of composite mood marking: The 736 non-Pama-Nyungan languages of northern Australia. Linguistic Typology, 9, 223–268. 737 Watahomigie, Lucille J, Bender, Jorigine, & Yamamoto, Akira Y. 1982. Hualapai reference grammar. 738 American Indian Studies Center, UCLA.