COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS IN LAKOTA: A ROLE AND REFERENCE GRAMMAR ANALYSIS Jan Ullrich The Language Conservancy In Lakota (Siouan) nouns (N) frequently occur adjacent to stative verbs (SV). Extant descriptions of Lakota grammar treat the <N1SV> as a syntactic compound in which the SV modifies the N. The present study offers a novel analysis that shows that the N and SV are uncompounded and that postnominal modification occurs only when the <N1SV> sequence is RP-internal, whereas in clause-final position it functions as a complex predicate. This analysis solves numerous outstanding issues from several areas of Lakota grammar including modification, modifier phrases, noun incorporation, inalienable possession, compounding, word formation, stress position, referentiality, and information structure. [Keywords: Lakota, complex predicate, coordination and cosubordination, modification, prosody] 1. Introduction.1 In Lakota (Siouan, ISO 639-3), nouns (N) frequently occur syntactically adjacent to stative verbs (SV). Extant Siouan literature treats 1 The data in this study originates in connected speech rather than from translational elicitation. About half of the data comes from a text corpus based in narratives and dialogues recorded primarily by the author between 1992 and 2019 from several hundred native speakers in Lakota communities in North and South Dakota. The other half originates from older Lakota texts collected before 1950s, namely from collections transcribed by Ella Deloria. Sources of data are indicated with the following abbreviations: DT (Deloria 1932), EDT (Ella Deloria Texts in the Boas Collection), BT (Buechel’s text collection, 1978), BBBJ (stories recorded from Ben Black Bear Jr., between 2004 and 2019), CHE (Clara High Elk, narratives, 2008), CWE (Charlie White Elk, narratives, 2009), DBE (Dave Bald Eagle, narratives, 2007–8), DTA (Delores Taken Alive, narratives recorded between 2005 and 2019), DW (David West, narratives, 1992–2010), FFC (Frank Fools Crow, narratives, unknown date), FREH (Florine Red Ear Horse, narratives, unknown date), IEC (Iris Eagle Chasing, narratives and dialogues, 2003–19), JHR (Johnson Holy Rock, narratives, 2003–7), JKS (Jerome Kills Small, narratives, 2007), LGH (Lakota Grammar Handbook [Ullrich 2016]), ML (Mary Light, narratives, unknown date), NLD (New Lakota Dictionary [Ullrich 2008]) NSB (Neva Standing Bear, narratives, unknown date), SLH (Shirley Left Hand, narratives, 2014–19), RFT (Rudy Fire Thunder, narratives and dialogues, 1992–93), RTC (Robert Two Crow, narratives and dialogues, 1998–2019). A partial archive of the narratives can be found in the Lakota Language Forum (https://www.lakotadictionary .org/), although the archive is being gradually moved to the Lakota Language Consortium YouTube channel (see, for example, https://www.youtube.com/watch?vplV25mDW3oMA). I want to thank Ben Black Bear Jr. and Iris Eagle Chasing for their help with grammaticality judgment on a couple of sentences whose word order was experimentally shuffled. I also want to express my gratitude to two anonymous reviewers who provided thoughtful comments on the manuscript. I thank Robert Van Valin for comments on an earlier version of this study. [IJAL, vol. 86, no. 3, July 2020, pp. 407–446] © 2020 by The University of Chicago. All rights reserved. 0020-7071/2020/8603-0003$10.00 DOI 10.1086/708833 408 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS the <N 1 SV> structure as a syntactic compound involving a postnominal modification in which the SV is subordinate to the N. When the nominal component of <N 1 SV> is a body part, the structure is generally treated as noun incorporation or noun stripping. In this study I challenge these widely held notions and offer evidence that the post-nominal modification analysis is relevant only when <N1SV> is RP-internal, whereas clause-final <N1SV> structures constitute a complex predicate, including those with body-part Ns.2 Additionally, the present investigation offers a revision of the prosodic features of <N1SV> and aims to show that the complex predicate analysis helps to account for a number of outstanding issues pertaining to Lakota grammar, such as first syllable stress compounds unaccounted for by the Dakota Stress Rule and structures with Ns that are not linked to any core arguments. Lakota is a head-marking language in which syntactic rules frequently target morphological elements, and conversely, affixes can have scope over syntactic structures. This is one of the reasons why the present investigation is written within the Role and Reference Grammar (RRG) framework, which is particularly well suited for analyzing such interaction between syntax and morphology.3 This paper is organized as follows: 2 provides a typological introduction to the Lakota language. 3 discusses the traditional treatment of the <N1SV> structures and offers a novel syntactic analysis. 4 is a discussion of <N1SV> structures involving inalienable Ns, which offer important evidence for and insights into the CP analysis. 5 investigates coordination of SVs. 6 and 7 describe the syntactic functions of CPs lexicalized as Ns and SVs respectively. In 8, I challenge the broad consensus that the <N1SV> sequence is compounded and I offer a new analysis of its prosody. 9 shows how the complex predicate analysis of 2 ‘RP’ stands for ‘reference phrase’, which is the category of referring expressions that are typically headed by nominals and thus have traditionally been labeled as noun phrases or ‘NP’. See Van Valin (2008) for a detailed discussion. 3 The following abbreviations are used: 1 ‘first person, 3 ‘third person’, A ‘actor’, ADD ‘additive particle’, ADV ‘adverb’, AFF ‘affix’, ANIM ‘animate’, ARG ‘argument’, AV ‘Active Verb’, CAR ‘Compound Accent Rule’, CAUS ‘causative’, CLM ‘clause-linkage marker’, CNTR ‘contrastive’, CP ‘complex predicate,’ DAR ‘Dakota Accent Rule’, DECL ‘declarative’, DEF ‘definite’, DEM ‘demonstrative’, DER ‘derivative, derivation’, DET ‘determiner’, DM ‘derived modifier’, DSR ‘Dakota Stress Rule’, ECS ‘extra-core slot’, EMPH ‘emphasis’, H* ‘high tone of the pitch accent’, HSY ‘hearsay’, INAN ‘inanimate’, INDEF ‘indefinite’, INSTR ‘instrumental’, INTJ ‘interjection’, L ‘low tone of the pitch accent’, LOC ‘locative’, LSC ‘layered structure of the clause’, LSW ‘layered structure of the word’, MOD ‘modifier’, MP ‘modifier phrase’, MSP ‘male speaker,’ N ‘noun’, NEG ‘negation’, NI ‘noun incorporation’, NUC ‘nucleus’, PL ‘plural’, PP ‘postpositional phrase’, PRED ‘predicate’, PRO ‘pronoun’, PROP ‘proper (name)’, PSR ‘possessor’, R ‘reference phrase level’ (only a subscript in LSC), REDUP ‘reduplication / reduplicated syllable’, REST ‘restrictive’, RC ‘relative clause’, REFL ‘reflexive’, RP ‘reference phrase’, RPFP ‘reference phrase final position’, RPIP ‘RP initial position’, RRG ‘Role and Reference Grammar’, S ‘sentence’, SG ‘singular’, SOV ‘subject object verb’, SV ‘stative verb’, U ‘undergoer’, V ‘verb’, W ‘word’. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 409 <N1SV> makes it possible to account for certain constructions in which an N is not linked to any core arguments. 2. Typological introduction to Lakota. This introduction to Lakota is necessarily brief and discusses only features relevant to the present study. Lakota is a left-branching, consistently head-marking, and strictly verb-final SOV language. It has head-internal relative clauses, many polysynthetic features, and no case marking on reference phrases (RPs). The subject and object are obligatorily coded as core arguments on the verb, whereas RPs are syntactically optional. Thus the verb alone can constitute a complete clause, and subject and object can be represented twice: as obligatory core arguments and as optional RPs that are cross‑referenced to the arguments and occur inside the clause in extracore slots (ECS). Significant for the present study is the fact that Lakota has an active/stative case-marking system (split-intransitivity, split-S; see Merlan 1985), where some intransitive verbs take actor (nominative) coding and other verbs take undergoer (accusative) coding. The active/stative distinction is made only in first singular and second persons, while all other grammatical persons neutralize it. When used on transitive verbs, the stative affixes signal the undergoer. Third person singular arguments are never marked overtly, and third person plural objects are marked only when the object is animate, in which case the affix wičha‑ is used.4 Animate plural of all other arguments is indicated with the suffix ‑pi. SVs can be reduplicated to indicate inanimate plural, and this can be done on SVs functioning both as predicates and modifiers. It is still not well understood whether or under what conditions reduplication of SVs is obligatory. The inflectional morphology is prefixal, but in accordance with Lakota grammatical tradition (e.g., Buechel 1939; Boas and Deloria 1941; Rood and Taylor 1996; and Van Valin 2005) the term ‘infix’ is also used as there are numerous instances where the inflectional morpheme is inserted in stems that are not analyzable synchronically, and some of which are difficult to analyze even with diachronic or comparative data. Two morphophonological features with marginal relevance to the present investigation are word-final ablaut and word-final syllable truncation. Ablaut is lexical, and ablauting verbs are indicated with capital A in glossing and word lists. Ablauting verbs alternate a, e, and iŋ depending on what they precede. Truncation is a process in which the final vowel is deleted and the remaining obstruent undergoes various phonological changes. 3. Syntactic analysis of <N1SV>. In Lakota, SVs are property words that can function predicatively without a copula. Lakota Ns, too, can function predicatively, and when the context allows it, they can conjugate with the same 4 The terms ‘subject’ and ‘object’ are used only as descriptive shortcuts as they have no theoretical status in RRG. 410 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS set of affixes as those used on SVs. For example, both the N hokšíla ‘boy’ and the SV čík’ala ‘to be small/little’ can function predicatively, as illustrated in (1).5 (1) (a) Homákšila háŋl . . . ho-má-kšila háŋl boy-1SG.U-STEM when (data: JKS-1: 40:20) AUDIO FILE 1A When I was a boy . . . (b) Mačík’ala čha . . . ma-čík’ala čha 1SG.U-small so (data: DBE-2: 8:41) AUDIO FILE 1B I was small so . . . The main objective of this study is to investigate how to account for constructions in which an N is syntactically adjacent to an SV. Since both can function predicatively, what are their respective syntactic functions in an <N1SV> sequence? And how is the constituency of <N1SV> affected by their syntactic relationship to each other and to other constituents? There has been an overall consensus in the Lakota research literature that the <N1SV> sequence represents post-nominal modification. This was first addressed descriptively in Riggs (1893 [1977]:45) and Buechel (1939:94), both of whom discussed SVs as ‘adjectives’, and more formally in Boas and Deloria (1941:70), who made the following statement: “The adjective follows the noun and is subordinate to it. The adjective is identical with the neutral verb. As a verb it retains its independent accent, as adjective it loses it. Šúŋka kiŋ tȟ áŋka ‘the dog is large’ šúŋka-tȟ àŋka ‘large dog’.” The post-nominal modification treatment of <N1SV> is also used by, among others, de Reuse (1978, 1994:201); Chambers and Shaw (1980:327); Shaw (1980:44); Williamson (1984:41); Rood and Taylor (1996:8.2.1), who describe <N1SV> as compounds; and Ingham (2003:13).6 De Reuse (1994:201) analyzes compounded <N1SV> structures as noun incorporation (NI). Rankin (2002) speculates that <N1SV> structures involving body-part Ns can be interpreted as possessor raising, as a transitive construction, or as NI, and he draws important conclusions regarding split intransitivity based on the assumption that the third option might be correct. The present study provides arguments against all three of the proposed analyses. 5 The ability of nominal predicates to take argument affixes, as in (1a), is not restricted to Ns denoting genders or life stages. There is a plethora of Ns that logically take an animate argument, for instance, wówačhiŋye ‘to be a dependable person/thing’. In fact, any noun can be marked for any grammatical person when used, for instance, by children during pretend play time: if they want to say such things as ‘I am a cat’ or ‘You are a tree’, they can use the stative affixes with any noun. 6 The post-nominal modification analysis is also generally accepted by authors working on other Siouan languages, e.g., Graczyk (2007:5, 271) for Crow, Rosen (2016:317) for Hocąk, etc. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 411 The discussions of <N1SV> in extant literature are almost without exception based on examples in which <N1SV> sequences occur in isolation rather than within a sentence originating from connected speech. Therefore, to begin the revision of the broadly accepted post‑nominal modification treatment, I offer two examples of <N1SV> from authentic discourse, given in (2) and followed by a discussion. AUDIO FILES 2A and 2B (2) (a) Hokšíla mačík’ala éyaš líla taŋyáŋ wéksuye ʔ. (data: JKS-5: 4:33) hokšíla ma-čík’ala éyaš líla taŋyáŋ Ø-wé-ksuyA boy 1SG.U-small but very well INAN-1SG.A-remember I was a little boy, but I remember it very well. (b) Naháŋȟ čiŋ hokšíla čikčík’alapi héhaŋni . . . (data: SLH-1: 3:00) naháŋȟ čiŋ hokšíla Ø-čik-čík’ala-pi héhaŋni still boy 3.U-REDUP-small-PL back.then Back when they were still little boys . . . In (2a), the affix ma‑ codes the first singular on the SV čík’ala ‘to be small’. In (2b), the same SV contains the zero affix to signal third person and the suffix ‑pi indicating animate plural of the argument, whereas the reduplication of the SV signals inanimate plural of the boys’ bodies (which indicates that the property expressed by the SV is predicated to the boys individually rather than collectively). This is evidence that the post-nominal SV in both examples is marked for a syntactic argument, and consequently it cannot function as a modifier and must be a predicate. However, the property predicated on the argument ma‑ in (2a) is not čík’ala ‘to be small’ but rather hokšíla čík’ala ‘to be a small boy’. And in the same way, the property ascribed to the zero affix in (2b) is the entire <N1SV> sequence. Therefore it must be concluded that the <N1SV> in each sentence in (2) constitutes a complex predicate (CP). In this CP both the N and the SV contribute their own semantics, but they share a single syntactic argument coded on the SV member. Although this syntactic structure is pervasive in Lakota, it has not been previously described. One of the defining properties of Lakota CPs is that they are always clause-final, as shown in (2), which is in line with the consistently verb-final syntax of the language. Note that the <N1SV> structure in (2a) is followed by a sentence-final glottal stop, which occurs whenever a predicate is sentence-final (and also after some sentence-final clitics). The generally accepted post-nominal modification analysis of <N1SV> is possible only in situations where the <N1SV> sequence is internal to an RP, as shown in (3), where it is not clause-final, and it is modified by the definite article kiŋ ‘the’. Note that the argument of the sentence-final V iyáyA ‘to leave’ is coded for plural with the suffix ‑pi, whereas the SV čík’ala ‘to be small’ does not have this suffix, which is evidence that the SV functions as a modifier and not as 412 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS a predicate. Consequently, the <N1SV> sequence in (3) is an instance of postnominal modification rather than a RC. (3) Hokšíla čikčík’ala kiŋ hená (data: SLH-7: 5:00) AUDIO FILE 3 iyáyapi na . . . hokšíla čik-čík’ala kiŋ hená i-Ø-yáyA-pi na boy REDUP-small the those leave-3.U-leave-PL and Those little boys left and . . . It is only when the <N1SV> occurs RP-internally, as in (3), that the SV can function as a modifier. It should be noted here that even though CPs are always clause-final, not every clause-final <N1SV> is a CP, and not every non-clausefinal <N1SV> sequence is an RP (the details are beyond the scope of the present study; for a discussion, see Ullrich 2018). Also, contrary to the overall agreement in the extant literature, the <N1SV> sequences in (2) and (3) are not compounded. This is discussed in 8. Further evidence in support of the CP analysis comes from the scope of negation and from the position of intensifiers. The negation operator šni ‘not’ in (4a) has scope over both words, and neither of them can be negated independently. Similarly, in (4b), the intensifier líla ‘very’ modifies the entire CP, and it is obligatorily positioned to its left. (4) (a) Hokšíla mačík’ala šni yeló. (LGH: U.20) hokšíla ma-čík’ala šni yeló boy 1SG.U-small NEG DECL.MSP AUDIO FILE 4A I am not a little boy. (b) Líla wíŋyaŋ niwášte čha . . . (EDT-Aut-3A: 77) líla woman ni-wášte čha very woman 2SG.U-beautiful so You are a very beautiful woman, so . . . Note that placing líla ‘very’ directly before the SV would render the sentence ungrammatical. The examples in (4) are evidence that in this construction the N is a co‑predicate and that it does not function as an RP. In order for an N to function as an RP linked to the argument of a predicatively functioning SV, the N has to be followed by a word that separates it from the SV. The most common separators are definite articles (i.e., kiŋ and k’uŋ), although quantifiers, partitives, additive particles, and sometimes postpositions can also function as such separators.7 The examples in (5) illustrate this point by 7 ‘Separator’ is just a convenient descriptive shortcut for words that intervene between the N and SV. Most separators are operators, but some have syntactic functions (see Ullrich 2018). It seems that indefinite articles can function as separators only before eventive SVs, such as t’Á COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 413 contrasting a complex predicate, in (5a), and a sentence with <N1DET> 1 SV functioning as a simple predicate, in (5b). The example in (5c) shows a quantifier in the position of a separator, which allows the N to function as an RP. (5) (a) Hokšíla háŋske ʔ. (data: LGH: U.6) AUDIO FILE 5A hokšíla Ø-háŋskA boy 3SG.U-to.be.tall He is a tall boy. (b) Hokšíla kiŋ háŋske ʔ. (data: LGH: U.6) AUDIO FILE 5B hokšíla kiŋ Ø-háŋskA boy the 3SG.U-to.be.tall The boy is tall. (data: BT: 349) (c) Oyáte tóna t’ápi ʔ. oyáte tóna Ø-t’Á-pi people several 3.U-to.die-PL Several people died.8 Note that the N in (5a) is non-referential, which is characteristic of Ns functioning as co-predicates in CPs. When Ns are RP-internal, as in (5b) and (5c), they can be referential or non-referential depending on the choice of determiners, context, and other factors.9 At this point it is important to say that although the Ns in (5a) and (5b) each have a different syntactic function, the subject of the two constructions is identical; it is the zero affix. In (5a), the N hokšíla ‘boy’ is a co‑predicate that shares the argument with the SV, whereas in (5b), the optional RP hokšíla kiŋ ‘the boy’ is linked to the argument at the clause level. This observation is important, because it has consequences for information structure in that in (5a) the argument is the sentential topic and the CP is the comment, whereas in (5b), the sentential topic is represented by both the argument and the RP linked to it, leaving the SV alone to express the comment. This becomes more apparent when the argument is expressed overtly (with an affix that has a phonological form), as shown in (6). In (6a), the first singular undergoer affix ma‑ is the subject and constitutes the sentential topic, whereas in (6b), the affix ma‑ is linked to the RP, making the latter a part of the topical ‘to die’, as in Ȟ tálehaŋ wíŋyaŋ waŋ t’é ‘A woman died yesterday’. SVs that indicate a property generally cannot function predicatively after indefinite RPs, except when the SV is a secondary predicate or when it occurs in an RC. 8 Since numerals take the stative affixes for person marking and function as stative verbs, they too have been commonly treated as modifiers when they occur post-nominally (e.g., Graczyk 2007:287). In reality, RP‑external non-truncated numerals function as secondary predicates, as detailed in Ullrich (2018). 9 For a discussion of referentiality in Lakota, see Ullrich (2018:2.7, 3.4). 414 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS information. The sentence in (6b) does not lend itself to a felicitous English translation, but it is a common construction in Lakota, and in fact, it can appear with any grammatical person, including the first singular and plural. (6) (a) Wíŋyaŋ matȟ áŋka háŋl . . . (data: EDT-Aut-4: 40) wíŋyaŋ ma-tȟ áŋka háŋl woman 1SG.U-be.big when When I was an adult woman . . . (b) Wičháša kiŋ lé omášteke ʔ. (data: EDT-FT: 21) wičháša kiŋ lé o-ma-štekA man the this be.peculiar-1SG.U-stem This man here, I am peculiar. Before I show the constituent projection of the constructions in (5a) and (5b), I should discuss one of the unique features of RRG crucial for the syntactic analysis of this type of complex predication. RRG recognizes three, rather than two, types of clause linkages: in addition to coordination and subordination, there are also constructions that are a kind of dependent coordination, in which clauses of equivalent status are joined together in a coordinate-like relation but share some grammatical category, for example, tense or evidentials. This dependent coordination is termed cosubordination, and it is a linkage in which two clauses are codependent. Furthermore, cosubordination can occur not only at the clause level but also at the level of the core and the nucleus. Cosubordination at clausal level is distinguished from coordination by clausal operator dependence. Thus it involves operator dependence between the units, unlike coordination, but not embedding, unlike subordination. Cosubordination of subclausal units (cores or nuclei) is characterized by operator dependence at the level of linkage (for details, see Van Valin 2005). CPs composed of <N1SV> show negation operator dependence at the NUC level because the words cannot be negated independently, as evidenced in (4a). This indicates that Lakota CPs involving <N1SV> are linked via nuclear juncture with a cosubordinating nexus (i.e., NUC cosubordination), which means that the N and SV are codependent. This is reflected in the constituent projection in Figure 1a, which shows the juncture at the NUC level. The layered structure of the word (LSW) allows visualizing the fact that the argument occurs at the level of the core of the word (COREW) rather than inside the core of the clause, which is in accord with the head-marking nature of Lakota.10 The argument is shared with the N via the nuclear juncture. This is different in Figure 1b, where the ARG is not shared with the N, because here the N functions as the head of an RP. The 10 For an account of RRG’s word-level nodes within the layered structure of word, see Van Valin (2013). COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 415 FIG. 1—(a) <N1SV> as a complex predicate. (b) SV as a simple predicate. RP is linked to the argument at the clause level, showing that the RP is dependent on the SV. Further pieces of evidence for the codependency analysis shown in Figure 1a are provided in most of the following sections. For comparison, Figure 2 offers the constituent projection of (3), where the SV functions as an RP-internal post-nominal modifier. Note that the SV is in the periphery of the NUCR, which makes it an ad-nuclear modifier (i.e., a postnominal modifier). The post-nominal modification function of SVs has been broadly recognized in the Siouan literature, but there are some properties of this construction FIG. 2—<N1SV> as an ad-nuclear modifier 416 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS that have been under-described or misunderstood, in particular the role of determiners.11 It should be noted here that there cannot be two or more RP-internal modifiers coordinated with the coordinator na ‘and’. This is a consequence of the fact that SVs function predicatively when they occur before conjunctions. Two syntactically different ways to ascribe more than one RP-internal property are discussed in 5, and another one is mentioned in 9. It needs to be added that there are SVs that can function only as simple predicates and not as members of CPs or as RP-internal modifiers. Examples of such verbs are watúkȟ a ‘to be tired’, čhaŋzékA ‘to be angry’, ípuzA ‘to be thirsty’, t’Á ‘to be dead, to die’, and possibly other SVs. The analysis of this restriction is beyond the scope of the present study. Regarding the CP composition, it should be mentioned that the first member of a CP can be any type of a derived N (in addition to simple Ns). Moreover, given the fact that Lakota Ns can function predicatively, it is not surprising that there are complex predicates composed of N1N. These occur very rarely and primarily in traditional personal names and geographical names. An example is in (7a), where the N máza ‘iron’ functions as the second co-predicate that takes the argument coding. This is contrasted with the example in (7b), where the same N functions as an incorporated pre-modifier. (7) (a) Čhaŋté Máza. (b) Máza-Čhaŋtè. čhaŋté Ø-máza máza-Ø-čhaŋtè iron-3SG.U-PSR.heart heart 3SG.U.PSR-iron His Heart Is Iron. He is (a person) of an Iron Heart. (free translation: He Has an Iron Heart.) Additional examples of CPs composed of N1N are Čhaŋté Pȟ éta ‘His Heart Is Fire’, Tȟ ašúŋke Wakíŋyaŋ ‘His Horse Is Thunder’, Wakíŋyaŋ Pȟ éta ‘Fire Thunder’ (literally: ‘A Thunder is Fire)’, and Blé Čháŋ ‘Wood Lake’ (literally: ‘A Lake is Wood)’. The Lakota CP construction with <N1SV> is different from various other types of complex predicates in other languages in that both of the co-predicates 11 As stated in 3, an <N1SV> sequence involves a post-nominal modification only when it constitutes an RP. Many RPs formed by <N1SV> are marked with a determiner, but not all <N1SV> sequences followed by a determiner are RPs, because they can also function as relative clauses or clefts. Conversely, there are <N1SV> sequences that constitute an RP even though they are not modified by a determiner, which is especially true when the nominal member is uncountable or notionally plural. De Reuse uses the term ‘Noun Stripping’ for RPs that are not modified with a determiner, but in Ullrich (2018:98–99) I argue against this terminology, because unmarked RPs are void of determiners due to the fact that they are plural or uncountable. In consequence, such RPs are not actually stripped of any elements, and the Noun Stripping is unsubstantiated with respect to Lakota RPs. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 417 contribute their own semantics, that is, neither one is a light verb, a linking verb, or a semantically impoverished verb.12 Understanding the NUC cosubordination analysis of CPs discussed in this section is essential in accounting for a number of unresolved issues of Lakota morphosyntax. Several of these will be discussed in the following six sections. 4. Complex predicates with body-part Ns. One of the outstanding problems in Siouan research literature is the syntactic analysis of body-part Ns adjacent to SVs. For instance, Rankin (2002) raises a question about the syntactic analysis of the sentence in (8): (8) Hí mayázaŋ ʔ. hí ma-yázaŋ tooth 1SG.U-to.hurt My tooth hurts. (data: Rankin 2002, but also occurring in the text corpus) He proposes that it could be accounted for with one of the three constructions cited below (Rankin 2002:22–23). (a) ‘A/The tooth hurts me.’ In this case hí is an ordinary noun subject, 1SG ma is the direct object, and the verb yazáŋ ‘hurt’ is in its unmarked third person. (b) ‘My tooth hurts.’ 1SG ma is the raised possessor of the subject ‘tooth’, and the verb is a third person sg form. Possessor raising is very common in Siouan languages. (c) ‘I (have a) tooth-ache.’ In this case hí is not the subject but rather an incorporated noun, and ma is a genuine stative subject that is infixed in a complex first person verb form. Rankin (2002:22) states that “if the third choice above is correct, then incorporation takes on greater significance in Siouan,” and he makes the following assertion: Utterances of this type are or were susceptible to all three labeled bracketings in several Siouan languages, and it is unclear whether this was originally an intransitive verb or a transitive verb with a body-part subject and a pronominal object that was later reinterpreted as a stative subject. So this particular verb, with its inherent ambiguities, may have been the “foot in the door” by which other active verbs could be reinterpreted as stative if they had experiencer subjects. (Rankin 2002:23) The noun incorporation analysis of (8) is used also by de Reuse (1994:202, 212–13) and Cumberland (2005:237). De Reuse (1994:202) states that “[e]ven 12 It is worth mentioning that unlike SVs, active verbs (AV) do not form CPs when they immediately follow an N. Instead, adjacent N1AV generally form a secondary predicate construction, which is a syntactically less tight construction in which the two co-predicates are connected via core-cosubordination. For a detailed discussion of secondary predication in Lakota, see Ullrich (2018). 418 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS though Stative verbs typically have one argument, NI has actually increased the valence of the verb here, so that the Stative verb . . . has two arguments, one being the Stative pronominal prefix ma‑ ‘1st person’, and the other the incorporated noun.” I propose that in reality the structure in (8) is a complex predicate (CP) like that in (5a). Consequently, the body-part N is not an RP and is non-referential. The CP analysis of (8) accounts for the fact that an intransitive verb has a syntactic relationship with two semantic participants, the body part and its possessor; the body part is a co-predicate of the SV, whereas the possessor is the argument of the SV. The argument is, therefore, shared between the N and the SV. Since body parts are inalienable possessions in Siouan languages, the shared argument is automatically interpreted as the possessor of the body part. CPs with body-part Ns do not lend themselves to felicitous English translations. The sentence “I am (a possessor) of a hurting tooth” probably reflects the Lakota structure in (8) more closely than other translations, but it is not a possible English sentence. Free idiomatic translations, on the other hand, often suggest structural or semantic properties that are not part of the original construction. For instance, “I am tooth-hurting” implies noun incorporation or compounding, neither of which is present in (8). This is a good example of how translations can often be misleading and suggestive of morphosyntactic properties not exhibited in the original. In addition to being used as co-predicates in <N1SV>, body-part Ns can function as RPs linked to the argument of an SV (i.e., in a simple predicate construction). The examples in (9) provide contrastive data showing the difference between a body-part N functioning as a co-predicate, in (9a), and an N that is internal to an RP, in (9b). Notice that while the first singular undergoer affix ma‑ is the subject in both sentences, they differ with respect to information structure: the sentential topic of (9a) is the first singular subject, whereas the sentential topic of (9b) is the subject together with the body-part RP linked to it. Consequently, (9a) is a proposition about the possessor (i.e., ‘I am a long-haired person’), while (9b) is a statement about the possessor’s hair (i.e., ‘My hair is long’). (body part as a co-predicate in a CP) (9) (a) Pȟ ehíŋ maháŋskaska ʔ. pȟ ehíŋ ma-háŋskA-ska hair 1SG.U.PSR-long-REDUP My hair is long. (data: LGH: U.156) AUDIO FILE 9A (Literally: ‘I am long-haired’ or ‘I am (a person) of long hair.’) (b) Pȟ ehíŋ kiŋ maháŋskaska ʔ. (body part as an RP) pȟ ehíŋ kiŋ ma-háŋskA-ska hair DEF 1SG.U.PSR-long-REDUP My hair is long. (data: LGH: U.156) AUDIO FILE 9B COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 419 Note that the reduplication of the SV in (9) signals the plural number of the N pȟ ehíŋ ‘hair’.13 Both of the constructions in (9) occur very frequently in corpus data. The fact that they express topically different propositions is illustrated in (10), which comes from a story about the narrator’s first flight in an airplane and in which the body part functioning as an RP is used to further emphasize where the pain occurred. AUDIO FILE 10 (10) Núŋǧ e mayázaŋ, áta núŋǧ e kiŋ mayázaŋ (data: DW-1: 10:33) čhaŋkhé . . . núŋǧ e ma-yázaŋ áta núŋǧ e kiŋ ma-yázaŋ ear 1SG.U.PSR-hurt intensifier ear the 1SG.U.PSR-hurt čhaŋkhé and so I had an earache, my ears really hurt and so . . . Strong evidence in support of the CP analyses lies in <N1SV> structures involving kinship terms, which, like body parts, are inalienable Ns in Lakota. Consider the examples in (11), where the quantifier óta ‘many’ functions as a stative co-predicate. Notice that in (11a) and (11b) the plurality of the children is indicated exclusively by the semantics of the quantifier, whereas the animate plural suffix ‑pi, which one might expect to be included, is not present. This is because the suffix ‑pi can signal only the plurality of the single argument in this CP, which is the parent (i.e., the ‘possessor’) of the children rather than the children themselves. A version with a plural argument is given in (11c) for comparison. (11) (a) Čhiŋčá óta ʔ. (data: EDT-Leg-10: 3) čhiŋčá Ø-óta children 3SG.U.PSR-many He has many children. (Literally: ‘He is (a parent) of many children.’) (data: DTA-8: 22:00) (b) Čhiŋčá maóta ʔ. čhiŋčá ma-óta children 1SG.U.PSR-many I have many children. (Literally: ‘I am (a parent) of many children.’) 13 It is worth mentioning here that AVs, too, can occur in both N1AV and RP1AV constructions (e.g., napé mayúze ‘he held me by the hand’ vs. napé kiŋ mayúze ‘he held my hand’). 420 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS (c) Čhiŋčá ótapi ʔ. (data: FREH-2: 0:25) AUDIO FILE 11C čhiŋčá Ø-óta-pi children 3.U.PSR-many-PL They had many children. (Literally: ‘They were (parents) of many children.’) The lack of ‑pi in (11a) and (11b) is evidence that čhiŋčá ‘his/her children’ is not an argument of the SV but rather its co-predicate. By contrast, the presence of ‑pi in (11c) indicates that the possessor (i.e., the parents) is the shared argument in this intransitive construction.14 For a complete account of the linkage of possessed Ns, we need to discuss examples in which the possessor is expressed overtly with an RP. In (12a), the possessor, represented by the proper name Hakéla, is an RP linked to the single argument of the CP. In (12b), on the other hand, pȟ ehíŋ kiŋ ‘the hair’ functions as an RP (which is licensed by the determiner kiŋ separating the N from the SV), and the possessor is internal to this RP. (12) (a) Hakéla pȟ ehíŋ šikšíčela ʔ. Hakéla pȟ ehíŋ Ø-šik-šíčA-la Hakéla hair 3SG.U.PSR-bad-REDUP-REST Hakéla’s hair is poor. (data: DT: 16-6) (Literally: ‘Hakéla is of poor hair.’) (b) Hakéla pȟ ehíŋ kiŋ eyá šikšíčelake č’éyaš . . . (data: DT: 11-8) Hakéla pȟ ehíŋ kiŋ eyá Ø-šik-šíčA-lakA Hakéla hair DEF oh.well 3SG.U.PSR-bad-REDUP-EMPH č’éyaš but Hakéla’s hair is, well, rather poor, but . . . Figures 3a and 3b show the constituent projections of (12a) and (12b) respectively. In Figure 3a, the proper name Hakéla functions as an RP linked to the argument of the CP in which pȟ ehíŋ ‘hair’ is one of the co-predicates. In Figure 3b, on the other hand, the possessor, Hakéla, is dependent on the N pȟ ehíŋ 14 The same evidence can be drawn from constructions involving alienable possessions, which participate in analogous structures except for the addition of the possessive prefix tȟ a‑ on the N. Compare the propositions about the owner (where the suffix ‑pi signals exclusively the plural of the owner)—Tȟ ašúŋke Óta ‘He is an owner of many horses’ and Tȟ ašúŋke Ótapi ‘They are owners of many horses’—with the propositions about the horses (where ‑pi on the predicate signals the plural of the horses and ‑pi on the N pluralizes the possessor): Tȟ ašúŋke kiŋ ótapi ‘His horses are many’ and Tȟ ašúŋkepi kiŋ ótapi ‘Their horses are many’. However, whenever a prefix indicating first person or second person is added to a possessed N, the latter generally functions as an RP. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 421 FIG. 3—(a) A body-part N as a co-predicate. (b) A body-part N as an RP. ‘hair’ and internal to the RP Hakéla pȟ ehíŋ kiŋ ‘Hakela’s hair’, which is then linked to the argument of the simple predicate. This contrast is enabled by the absence or presence of the separator kiŋ ‘the’. (Note that eyá ‘well’ is an interjection and has no impact on the syntactic structure.) Another aspect relevant for the CP analysis of <N1SV> lies in the position of the locative prefix i‑, which can occur either on the N or the SV, as shown in (13a) and (13b) respectively. The fact that the locative can be prefixed to the N represents strong evidence in favor of treating the N as a co-predicate and not as an RP. (13) (a) Ináta mayázaŋ ʔ. (IEC-7: 11:40, also EDT-Bo: 19: ithézi yazáŋ) i-náta Ø-ma-yázaŋ AUDIO FILE 13A LOC-head INAN-1SG.U.PSR-to.hurt I have a headache from it. (b) Natá imáyazaŋ ʔ. (IEC-7: 12:30) AUDIO FILE 13B natá i-Ø-má-yazaŋ head LOC-INAN-1SG.U.PSR-to.hurt I have a headache from it. A similar argument can also be made with respect to instrumental prefixes, which, too, can occur before either one of the co-predicates, as shown in (14). Whereas the placement of the instrumental prefix yu‑ before the SV, as in (14c), is possible, it is significantly less frequent when compared to the position before the N, shown in (14b). 422 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS (14) (a) Čhaŋté mawášte ʔ. (data: EDT-Col-3: 69) čhaŋté ma-wášte heart 1SG.U.PSR-to.be.good I am glad. (Literally: ‘I am of good heart.’) (data: BBBJ-3: 5:30) Audio 14B (b) Mayúčhaŋte wašté ʔ. ma-Ø-yú-čhaŋte wašté 1SG.U.PSR-3SG.A-INSTR-heart good He made me glad. (data: RTC-16: 36:00) Audio 14C (c) Čhaŋté mayúwašte ʔ. čhaŋté ma-Ø-yú-wašte heart 1SG.U.PSR-3SG.A-INSTR-good He made me glad. Instrumental prefixes change CPs into transitive verbs, and the causative meaning of the instrumental prefix has scope over both members of the <N1SV> sequence, which is accounted for with the nuclear cosubordination juncture. One might argue that the leftmost position of the prefix could result from the fact that čhaŋté wašté ‘to be glad’ is a lexicalized expression, but in reality, we find instrumentals prefixed to just about any random CP. An example is shown in (15), which is a sentence from a story about a cultural hero who transforms himself into a large hawk (hence the instrumental prefix yu‑ is changed into the reflexive form iglú‑). (data: EDT-Col-3, sentence 266) (15) Iglúčhetaŋ tȟ áŋka ʔ. Ø-igl-(y)u-čhetáŋ tȟ áŋka 3S.A-REFL-INSTR-hawk large He transformed himself into a large hawk. The fact that instrumental and locative prefixes have scope over both members of the <N1SV> sequence is evidence that it constitutes a CP. At the same time, the possibility of placing the prefix on either of the words supports the cosubordination analysis, because it is evidence that the co-predicates are of equal status in the construction. This would not be possible if there were a head verb and a subordinate verb. Additionally, the possibility of placing these prefixes before the SV member of the CP indicates that these <N1SV> sequences are not phonetically tight, contrary to the general consent in the extant Siouan literature. The phonological tightness of the <N1SV> CPs is addressed in 8. 5. Coordinating stative verbs. Ascribing two or more properties to a single body-part N or ascribing a single property to two or more body-part Ns represents an area that has received little to no attention in the research literature. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 423 Nonetheless, it is a feature of the grammar that can provide interesting insights into inalienable Ns as well as into the properties of complex predication and coordination, which is the reason why it is included in this study.15 When two or more SVs are conjoined with the coordinator na ‘and’, each of them heads its own clause, as shown in (16). The evidence lies in the obligatory pluralization of the SVs when their argument is plural and animate, as in (16b). (16) (a) Lé-aŋpètu kiŋ tȟ áŋka na wašté ʔ. lé-aŋpètu kiŋ Ø-tȟ áŋka this-day the INAN-great (data: EDT: Del-Leg-5, sentence 74) na and Ø-wašté INAN-good This day is great and good. (Literally: ‘This day is great, and it is good.’) (data: BT: 222) (b) Čhépapi na waštépi ʔ. Ø-čhépa-pi na Ø-wašté-pi 3SG.U-fat-PL and 3SG.U-good-PL They were well nourished and handsome. (Literally: ‘They were well nourished, and they were handsome.’) In (16a), the argument of the SV tȟ áŋka ‘great’ is linked to the RP lé-aŋpètu kiŋ ‘this day’. The second SV, on the other hand, has no intra-clausal relationship with the RP, so the understanding that it ascribes a property to the RP in the first clause is based solely on the fact that the argument in each of the two clauses is coded for the same grammatical person. The above observation has important implications for coordination of SVs ascribing properties to body-part Ns, because if the body-part RP inside the first conjoined clause is not linked to the SV heading the second conjoined clause, then the subject of the first clause is linked to the body part as its possessor but the subject of the second clause is not linked to it at all. As a consequence, the SV in the second clause cannot ascribe a property to the body part. That this is the case is shown in (17a), which native speakers judge as nonsensical. The SV škópA ‘to be curved or crooked’ is not a logical property 15 A brief mention can be found in Rudin (2016:379), who states that Omaha SVs have to be in separate coordinated clauses in order to describe a single participant (although she does not discuss inalienable nouns). Cumberland (2005:386–88), working on the closely related Assiniboine Nakhóta language, states that “multiple stative verb modifiers in an NP are rare in spontaneous speech. In fact, none occur in the narrative or language circle tests used for this study. All such examples of this type are elicited, and speakers have difficulty producing them, frequently giving conflicting responses. Consequently, I have not been able to find a pattern in the construction of NPs of this type, despite coming back to the question several times over a period of three years.” The elicited examples that Cumberland provides are contradictory to the findings from corpus data offered in this section. 424 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS of a person, but since it is heading the second coordinated clause, it is not linked to the body-part RP, and consequently it ascribes property to the person rather than the body part. If we swap the order of the clauses, as in (17b), the sentence becomes logical, because škópA ‘to be crooked’ is a possible description of a nose and tȟ áŋka ‘to be large’ can be a description of a person. (17) (a) * Pȟ asú kiŋ matȟ áŋka na maškópe ʔ. pȟ asú kiŋ ma-tȟ áŋka na ma-škópe nose the 1SG.U.PSR-large and 1SG.U-crooked (Intended: ‘My nose is large and crooked.’) (Literally: ‘My nose is large, and I am crooked.’) (data: BBBJ: gram. judg.) (b) Pȟ asú kiŋ maškópiŋ na matȟ áŋka ʔ. Pȟ asú kiŋ ma-škópA na ma-tȟ áŋka nose the 1SG.U.PSR-crooked and 1SG.U-large My nose is crooked, and I am big. (data: BBBJ: gram. judg.) The fact that (17a) is considered illogical is evidence that body-part RPs linked to the argument of the SV’s heading the first coordinated clause are not linked to the argument of the SV’s heading the second coordinated clause. The situation is different when the SVs in both of the coordinated clauses are coded with the zero affix, which is not restricted to coding the third person singular but can also signal an inanimate argument. This is illustrated in (18), which is a grammatical and perfectly logical statement. (18) Pȟ asú kiŋ tȟ áŋka na škópe ʔ. (data: IEC) pȟ asú kiŋ Ø-tȟ áŋka na Ø-škópe nose the 3SG.U.PSR-large and INAN-crooked His nose is large and crooked. (Literally: ‘His nose is large and it is crooked.’) The grammaticality of (18) can be accounted for through the fact that the zero affix can signal both third person singular and inanimate arguments. Thus the RP pȟ asú kiŋ ‘his nose’ is syntactically linked to the possessor coded with the zero affix on the SV tȟ áŋka ‘his is large’, and it has a semantic link with the zero coded inanimate argument of the SV škópe ‘it is crooked’. In other words, the two conjoined clauses in (18) have grammatically different subjects even though notionally they have the same referent. Only the zero affix makes it possible for sentences like (18) to be logical, whereas any of the affixes with a phonological form renders such sentences nonsensical, as was shown in (17a). Additionally, the ability of the zero affix to code both the possessor and the INAN enables the coordination of a clause headed by a CP involving a body-part COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 425 N with a clause headed by an SV functioning as a simple predicate, as illustrated in (19). (19) Pȟ ehíŋ sapsápiŋ na akhíšokiŋ na . . . (Data: EDT-Col-4: 322) pȟ ehíŋ Ø-sap-sápA na akhí-Ø-šokA hair 3SG.U.PSR-black-REDUP and thick-INAN-thick His hair was black and thick and . . . (Literally: ‘He was of black hair and it was thick.’) In (19), the argument of the first coordinated clause signals the third singular possessor (‘he’), whereas the argument of the second clause signals inanimate undergoer (‘it’). There is no syntactic relationship between the body-part N, which functions as a co-predicate in the first clause, and the SV heading the second clause. Nonetheless, the semantic linking of different referents with the zero affix allows (19) to be a well-formed sentence. But replacing the zero affixes with any of the overt affixes would render the sentence illogical. The construction in (19) is interesting with respect to referentiality; the bodypart N has a syntactic relationship to the first SV, but its relationship with the second SV is only semantic. Consequently, the non-referential body part of the CP becomes referential because arguments are by default referential. The construction in (18) seems more logical, because the RP-internal N is referential, and so its referentiality in the second clause does not change. Sentences like (19) are very rare in corpus data, and it is not entirely impossible that this construction is idiosyncratic or the result of omitting the definite article by the scribe.16 It has long been generally agreed that the conjunction na ‘and’ requires the same subject in the conjoined clauses; see, for example, Chafe (1976) and Dahlstrom (1982). However, the examples in (18) and (19) indicate that na ‘and’ can conjoin clauses with different subjects, because even though both SVs have the same affix, they do not have the same subject. In fact, clauses coordinated with na and each having a different overtly marked argument are quite common in corpus data.17 The constituent projections of (18) and (19) are shown in Figures 4a and 4b. 16 Only a couple of instances of the construction shown in (19) can be found in the corpus, and they both originate from older texts for which there is no audio recording. Thus it can be the case that the lack of kiŋ ‘the’ in these sentences is a transcription error, because kiŋ ‘the’ is often pronounced as a g in fast speech, where it is often hard to hear, especially for an inexperienced scribe. However, the present analysis does account for the construction without kiŋ ‘the’, and it is therefore possible that its low frequency in corpus data is a feature of the limited size of the corpus. 17 The same subject requirement on the Hocąk counterpart of the conjunction is reported by Johnson et al. (2016:352), whereas the Assiniboine coordinating conjunction híŋk allows different subjects (Cumberland 2005:411). Also there are some indications that the Lakóta na ‘and’ may also function as a cosubordinator under certain conditions. 426 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS FIG. 4—(a) Body-part RP syntactically linked to the zero affix on the first SV and semantically linked with the zero affix of the second SV. (b) Zero affix allowing CP coordinated with a clause headed by SV to ascribe two properties to a single body part. There are a number of ways to ascribe two different properties to a single noun when the subject is in a grammatical person other than the third singular. One of them is illustrated in (20), which shows two CPs with third plural subject conjoined with na ‘and’. This entails repeating the body-part N, which in turn ensures that the argument of each clause is shared with the same body-part N co-predicate. AUDIO FILE 20 (20) Pȟ ehíŋ háŋskaskapi naíŋš pȟ ehíŋ waštéštepi ʔ. (data: FREH-19: 1:16) pȟ ehíŋ Ø-háskA-ska-pi na-íŋš pȟ ehíŋ hair 3SG.U.PSR-long-REDUP-PL and-CNTR hair Ø-wašté-šte-pi 3SG.U.PSR-good-REDUP-PL They had long and beautiful hair. (Literally: ‘They were (people) of long hair, and they were (people) of beautiful hair.’) Another way to ascribe two properties to a single body-part N is by compounding two SVs into a single phonological word. Lakota allows this via morphological pre-modification, that is, the first member of the compound modifies COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 427 the second member. This is shown in (21), which illustrates that compounded SVs can ascribe attributes to the participant via complex predication, as in (21a), or via simple predication, as in (21b). (21) (a) Pȟ ehíŋ zizí-waštèšte ʔ. (data: EDT-Col-4: para 33) pȟ ehíŋ zi-zí-Ø-waštè-šte hair yellow-REDUP-3SG.U.PSR-good-REDUP Her hair was beautifully yellow. (Literally: ‘She was of yellow-beautiful hair.’) (b) Pȟ ehíŋ kiŋ waštéšte-hàŋskaska ʔ. (data: Del-Col-4: para 138) pȟ ehíŋ kiŋ wašté-šte-Ø-hàŋska-ska hair the good-REDUP-3SG.U.PSR-long-REDUP His hair was nice and long. (Literally: ‘His hair was nice-long.’) Another common strategy of avoiding SV coordination is to change one of the SVs into a derived modifier (DM) using the suffix ‑ya. Such DMs can occur in combination with SVs or be serialized. For example, the intended meaning of the ungrammatical sentence in (17a) can be expressed by (22): (22) Pȟ asú kiŋ škobyá matȟ áŋka ʔ. (data: BBBJ) pȟ asú kiŋ škópA-ya ma-tȟ áŋka nose the crooked-MOD 1SG.U.PSR-large My nose is big and crooked. (Literally: ‘My nose is crooked big.’) Derived modifiers like škobyá ‘crooked’ in (22) function as RP-external modifiers, and depending on their semantics and on the syntactic structure they appear in, they have scope over the argument or the core, and in some constructions even over the RP.18 Additional phenomena related to the construction in (22) are discussed in 9. Coordinating two different body-part Ns in order to ascribe a single property to them is another area that provides further evidence in support of the CP analysis. This is shown in (23), which offers three different options of expressing such a concept. In (23a), there are two coordinated CPs, each involving the same body-part N. Compare this with (23b), which shows that omitting the first SV in an attempt to coordinate the Ns results in an ungrammatical sentence, because only the second N can now function as a co-predicate, whereas the first N has no syntactic relationship to the SV. It is, however, possible to coordinate the Ns when they function as RPs, as shown in (23c) and (23d). In both of these 18 For a detailed discussion of derived modifiers, see Ullrich (2018: chap. 5). 428 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS sentences, the additive particle kȟ ó ‘too’ is used as the separator, and the data show that kȟ ó can be used in concert with the conjunction na ‘and’, as in (23c), or without it, as in (23d). Both (23c) and (23d) are judged as ungrammatical without the additive particle kȟ ó. (23) (a) Líla niǧ é tȟ áŋka na natá tȟ áŋka kéyapi. líla niǧ é Ø-tȟ áŋka very belly 3SG.U.PSR-big Ø-kéya-pi 3.A-say.that-PL (data: NSB-4: 1:30) Audio 23A na and natá Ø-tȟ áŋka PSR.head 3SG.U.PSR-big They say he had a very big belly and a big head. (Literally: ‘They say he was of a very big belly, and he was of a big head.’) (b) * Líla niǧ é na natá tȟ áŋka ʔ. (grammaticality judg.: BBBJ) líla niǧ é na natá Ø-tȟ áŋka very belly and PSR.head 3SG.U.PSR-big (Intended: ‘He had a very big belly and head.’) (c) Pȟ ehíŋ na ištá kȟ ó zizí na . . . (data: EDT-Col-4: para 12) pȟ ehíŋ na ištá kȟ ó Ø-zi-zí na hair and eye ADD 3SG.U.PSR-yellow-REDUP and Her hair and eyes, too, were tawny, and . . . (d) Pȟ ehíŋ ištá kȟ ó sabsápiŋ na . . . (data: EDT-Col-4: para 318) pȟ ehíŋ ištá kȟ ó Ø-sab-sápA na hair eye ADD 3SG.U.PSR-black-REDUP and Her hair and eyes, too, were black, and . . . Sentences like those in (23a), (23c), and (23d) can occur with any grammatical person coded on the SV, which is not possible in the structures shown in (18) and (19). The data in (23) offers important confirmation for the CP analysis and its syntactic properties. It is worth noting that multiple properties can also be ascribed via stacked relative clauses, a construction beyond the scope of this study (for details see Ullrich 2016). The above discussion of coordinated SVs illustrated in (18)–(22) provides further insights into Lakota <N1SV> complex predication and RP-internal modification with SVs. Furthermore, it gives additional evidence in support of the non-configurationality of Lakota, as shown in Van Valin (1985, 1987), and against the claims that Dakotan languages are configurational, for example, in West (1999, 2003). COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 429 6. Complex predicates lexicalized as nouns. Since CPs are descriptions of entities, many of them become lexicalized as Ns and names of nominal classes. Examples are kȟ aŋsú-sutá ‘certificate’ (literally: “hard card”), mázawakȟ áŋ ‘gun’ (literally: “magical iron”), máza-ská ‘money’ (literally: “white metal”), wašíču-wakȟ áŋ ‘doctor’ (literally: “powerful white person”) and šiná-lúta ‘red blanket’. Such lexicalized CPs are variously written as a single word or two separate words, depending on orthographic conventions and the level of awareness of the lexicalization. To show that these nominal expressions originate in a syntactic construction involving two words, I am spelling them with a hyphen here. As a consequence of their nominal status, they can also function as nominal co-predicates with SVs, resulting in a structure that can be schematically shown as [[N [N1SV]]1[SV]]. Examples are in (24). (24) (a) Šúŋka-wakȟ áŋ wašté ʔ. (data: EDT-Bo: 95) šúŋka-wakȟ áŋ Ø-wašté horse(dog-magical) 3SG.U-good It is a good horse. (b) Máza-ská óta ʔ. (data: CHE-1: 21:19) Audio 24B máza-ská Ø-óta money(metal-white) INAN-much It is a lot of money. The constituent projection of (24a) is given in Figure 5. It shows that the NUC cosubordination connecting the <N1SV> complex predicate is at the phrase level, whereas the NUC cosubordination connecting the N and SV inside the lexicalized CP is at the word level. Unless lexicalized CPs are spelled as a single word or with a hyphen, the construction in (24) may appear to have two syntactically adjacent SVs, which Lakota does not allow. In reality, SV1 is connected via NUC cosubordination to the N at the word level, and the resulting N is connected through NUC cosubordination at the phrase level with SV2. The combination of the layered structure of the clause (LSC) and the layered structure of the word (LSW) allows us to visualize this effectively. Lexicalized CPs can also function as incorporated pre-modifiers. This is illustrated in (25). Note that this is a morphological construction in which the <N1SV> structure is compounded with the N it precedes, resulting in a reduction of stress on the N member.19 19 For more details on pre-modification in Lakota, see Ullrich (2018: chap. 8). 430 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS FIG. 5—A lexicalized CP functioning as an N member of a CP (25) (a) Šiná-lùta-čhuwìgnaka ʔ. šiná-lúta-Ø-čhuwígnaka blanket-red-INAN-dress It was a red-cloth dress. (b) Makȟ ásaŋ-Oyàte kiŋ hél uŋthípi ʔ. makȟ á-sáŋ-oyáte kiŋ hél clay-white-people the there (data: FFC: 15:00) Audio 25A (data: ML-2: 0:30) Audio 25B uŋ-thí-pi 1A-live-PL We lived in White Clay Community. The constituent projection of (25a) is provided in Figure 6. The lexicalized CP analysis also helps to account for certain traditional placenames that can have two different word orders, as in (26). In (26a), the N makȟ áska ‘white earth’ formed by the <N1SV> CP functions as an incorporated pre-modifier of the N blé ‘lake’. In (26b), on the other hand, the N makȟ á-ska ‘white earth’ functions as the second co-predicate in an <N1N> complex predicate (i.e., it is the same construction as in (7a)). (26) (a) Makȟ á-Ska-Blè Incorporated pre-modification: [N[N1SV]-N] makȟ á-ská-blé earth-white-lake White-Earth-Lake COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 431 FIG. 6—A lexicalized complex predicate functioning as a nominal pre-modifier (b) Blé Makȟ á-Ská Complex predicate: [N1N[N1SV]] blé makȟ á-ská lake earth-white Lake of White Earth (Literally: ‘A Lake Is White Earth’) Another example of this type of variation is observed in the name of the Minnesota River: (a) Mnisóta-Wakpà ‘Milky-Water-River’ and (b) Wakpá Mnisóta ‘River of Milky Water’. This section showed that the CP analysis allows us to account for Ns composed of <N1SV> and their function as the nominal member of <N1SV> CPs. 7. Complex predicates lexicalized as SVs. Whereas some CPs become lexicalized as Ns, as shown in the previous section, other CPs can be lexicalized as SVs. Examples of such complex SVs are oȟ ’áŋ-wašté ‘to be (a person) of good manners’ and wíŋyaŋ-wašté ‘to be a beautiful woman’ or ‘to be (a person) of feminine beauty’. Such complex SVs commonly function as simple predicates, as in (27a), and less frequently also as the second members of <N1SV> CPs, as in (27b), and they can even appear as RP-internal post-modifiers, as in (27c) and (27d). 432 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS (27) (a) Hokší-hakákta kiŋ [RP[N1det]1SV[N1SV]] (data: BT: 381) oȟ ’áŋ-waštépi ʔ. hokší-hakákta kiŋ oȟ ’áŋ-wašté-pi child-youngest the manner-3U-good-PL The youngest children had / were of good manners. (b) Wičháša oȟ ’áŋ-wašté ʔ. [N1SV[N1SV]] (data: EDT-Bo-43: 2) wičháša oȟ ’áŋ-Ø-wašté man manner-3SG.U-good He is a man of good manners. (c) Wičháša oȟ ’áŋ-wašté waŋ [RP[N1SV] 1 V] (data: BBBJ-9: 3:00) hiŋgnáye ló. wičháša oȟ ’áŋ-wašté waŋ hiŋgná-Ø-Ø-yA man manner-good a husband-3SG.U.3SG.A-CAUS ló DECL.MSP She married a man of good manners. (d) Wikȟ óškalaka wíŋyaŋ-waštéšte [RP[N1SV] 1 V] (data: DT: 26-2) óta thípi kiŋ hél . . . wikȟ óškalaka wíŋyaŋ-wašté-šte óta Ø-thí-pi young.women woman-beautiful-REDUP many 3SG.A-dwell-PL kiŋ hél the there There, where many young women of feminine beauty lived . . . The constituent projection of (27b) is given in Figure 7a, and it shows an NUC cosubordination at the word level, signaling that the SV is a lexicalized CP. An alternative analysis of this sentence is provided in Figure 7b, where oȟ ’áŋ wašté ‘to be (a person) of good manners’ is treated as a CP stacked inside a higher-level CP. This treatment employs a very tentative hypothesis that it might be possible for some CPs to function in this way without being lexicalized first. The constituent projection of (27c) is given in Figure 8, and it shows that the NUC cosubordination is in the word-level nucleus (NUCw) of the RP‑internal modifier rather than in its phrase-level nucleus (NUCM). This indicates that the SV is a lexicalized CP. The post-nominal (ad-nuclear) modification is licensed by the fact that the SV is RP‑internal. It is likely that non-lexicalized <N1SV> sequences can also function as RP-internal modifiers. For instance, we find sentences like the following: Wagnúka pȟ á šá waŋ él híyotaka čhaŋkhé . . . ‘A red headed woodpecker sat down on it and so . . .’. (data: DT: 7:11), in which the <N1SV> pȟ á šá ‘to be (a possessor) of red head’ is spelled as two independent words by Deloria COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 433 FIG. 7—(a) CP lexicalized as an SV, functioning as the second member of a complex predicate. (b) Stacked complex predicate. (1932), who otherwise had a tendency to over-apply compounding. Nonlexicalized <N1SV> functioning as RP-internal modifiers are connected via NUC juncture at the phrase level. More investigation is needed to determine how common this is in the corpus. FIG. 8—Lexicalized <N1SV> CP functioning as a complex RP-internal modifier 434 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS 8. Prosodic analysis of <N1SV>. In this section I offer a revision of the commonly accepted analyses of the prosodic properties of the <N1SV> constructions. As mentioned earlier, Boas and Deloria (1941:70) state explicitly that an SV loses its stress when it modifies an N. In addition, they show numerous minimal pairs contrasting in phonological tightness, as those in (28) (glosses added by me): (28) (a) čháŋ-wakȟ àŋ tree-holy a sacred pole (b) čhaŋwákȟ aŋ tree.holy (data: BD-1941: 70) sundance pole The minimal pair in (28) illustrates the distinction between two generally accepted types of compounding first categorized by Chambers (1978), who describes them as follows: Syntactic compounds, as in (28a), are compounds in which both members keep their stresses, but the stress on the second member is reduced. In lexical compounds, a single stress is assigned, as in (28b). Although this categorization is in line with the numerous minimal pairs provided by Boas and Deloria, there are various compounded and uncompounded constructions that cannot be accounted for using this categorization, as I will explain in this section. First, I will offer evidence against treating <N1SV> constructions like that in (28a) as compounded. Second, I will show that there are compounds that are arguably lexical and yet have both primary and secondary stresses. And third, I will discuss compounds that are syntactic in their nature and yet have a single stress. I begin with evidence from morphosyntax given in (29), which shows a word occurring between the N and the RP-internal SV. The intervening word is the intensifier líla ‘very’. This separation makes it obvious that the N and SV are not compounded. (29) Šúŋkawakȟ áŋ líla waštéšte k’eyá iyéwičhaya čhaŋkhé . . . k’eyá šúŋkawakȟ áŋ líla wašté-šte horse very good-REDUP INDEF iyé-wičha-Ø-yA find-3PL.U.ANIM-3SG.A-STEM (data: EDT-Bo-7: 3) čhaŋkhé and so He found some very good horses and so . . . Why is it, then, that there is a wide consensus that SVs functioning as postnominal modifiers form a compound with the Ns they modify? An important part of the answer to this question lies in the fact that Boas and Deloria (1941) based their description of the prosodic properties of <N1SV> on isolated instances of this structure rather than on its occurrences in texts. Had they observed the <N1SV> sequences in sentences originating from connected speech, COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 435 they would have likely found syntactic evidence against compounding, like that in (29), which shows sentences from texts transcribed by Deloria. Second, until now there has not been a study using prosodic data that would place the Boas and Deloria description under scrutiny, although some researchers have commented that the Boas and Deloria transcription may not be reliable with respect to compounding and secondary stress. For instance, de Reuse (1994:206), who, based on impression, observed different compounding patterns in his work with native speakers, states that “[t]here are typos and errors in the materials written by Boas and Deloria.” The first acoustic analysis of Lakota prosody is Mirzayan (2010), and even though his investigation is not concerned directly with <N1SV> structures, it provides important findings about Lakota pitch accent and the acoustic properties of the Lakota intonational phrase. His observations can be used to enhance our understanding of the <N1SV> structures. Mirzayan (2010:119) states the following: “The first, and most important, contribution to F0 drop in the intonational phrases in Lakota is the application of downstep at specific points. Phonetically, downstep lowers subsequent H* peaks inside phrasal units, establishing a new high level for the remaining part of the utterance at each point of application. In this manner, downstep causes the tonal space to contract in a cascading staircase as the utterance progresses.” (F0 is ‘fundamental frequency of the voice’ or ‘pitch contour’.) This process is illustrated schematically in Figure 9. In order to test whether the downstepping of the second H* peak within an intonational phrase applies to the <N1SV> constructions, I conducted pitch analysis (using Praat acoustic tools) on a large number of examples from audio recordings of connected speech. This resulted in much evidence in support of the assertion that the <N1SV> does not form a phonological word. An example is the pitch contour of (30), shown in Figure 10. FIG. 9—Schematic representation of two H* pitch accents. “The second accent is downstepped relative to first. The tonal space, defined as the amount of pitch range used above a baseline unit, contracts with the application of downstep” (Mirzayan 2010:119). 436 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS FIG. 10—Praat speech analysis of H* pitch downstep (30) Wakȟ áŋheža čík’ala ób égna (data: JHR01, 2005: 19:20) AUDIO 30 maŋké ʔ. wakȟ áŋheža čík’ala ób égna m-aŋkÁ child small with among 1SG-sit I sat among small children. Figure 10 shows that the H* pitch of the SV čík’ala ‘small’ is downstepped relative to the *H pitch of the N wakȟ áŋheža ‘child’, thus supporting Mirzayan’s description of a pattern occurring in intermediate intonational phrases (a mid-level prosodic phrase that is usually larger than a word but included inside an intonational phrase). The sequence can be coded as H* !H*, where ! is the symbol for downstep. In a discussion with Mirzayan, we came to the hypothesis that it might have been this downstep of the H* peaks that lead Boas and Deloria into thinking, based on impression, that the second downstepped member of the intermediate phrase had lost its independent stress. However, the H* peak downstep is a large (above word level, at the phrase level) prosodic and intonational phenomenon, and it is not equal to stress reduction. Due to the downstepped H* peak, it is not uncommon that lexicalized CPs are pronounced with a weaker stress on the second member, which is reflected in the traditional spelling of these sequences, such as čhaŋtéwašte ‘to be glad’ (literally: ‘to be of good heart’). However, even such lexicalized expressions are COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 437 variably pronounced with stress on the second member reduced (čhaŋtéwaštè) or full with downstepped *H peak (čhaŋté wašté). Another example is from a recording of wašíču wakȟ áŋ ‘doctor’ pronounced with clear independent stress on wakȟ áŋ despite the fact that this <N1SV> is a lexicalized N. Depending on the tempo of speech and other prosodic aspects, an SV in post-nominal position may be perceived as having a slightly weaker stress, but as mentioned earlier, this is a matter of intonation rather than stress reduction.20 The pitch contour analysis presented above is evidence against the broad consensus that the syntactic <N1SV> structures are compounded. This is in addition to the evidence from syntax and morphology presented in (12) and (14) respectively. However, not all syntactic <N1SV> structures have the same prosodic properties as those illustrated in Figure 10. In particular, when the N is a monosyllabic word and the SV is stressed on the first syllable, the CP construction is stressed on the first syllable and compounded, as shown in (31). (data: BBBJ, CWE-1: 3:25) (31) (a) Ípuze ʔ. í-Ø-púzA mouth-3SG.U-to.be.dry AUDIO 31A He is thirsty. (Literally: ‘He is dry-mouthed.’) (b) Ínuŋpa ʔ. í-Ø-núŋpa mouth-INAN-two (data: NLD, EDT-Rig: 200) AUDIO 31B It is double barreled. (Literally: ‘It is two-mouthed.’) (data: DT-23: 7) (c) Híŋšiče ʔ. híŋ-Ø-šíčA fur-3SG.U-to.be.bad It has poor fur. (Literally: ‘It is bad-furred’ or ‘It is (a possessor) of a bad fur’.) The word-initial stress of these compounds cannot be accounted for by the Dakota Stress Rule (DSR), which was formalized by Shaw (1980, 1985), building upon, and extending, the prior work by Carter (1974) and Chambers (1978), all of whom worked primarily with the data from Boas and Deloria (1941). The DSR (originally named Dakota Accent Rule by Shaw) puts an accent on the 20 It is not unlikely that RP-internal <N1SV> structures have somewhat different prosodic properties from <N1SV> structures constituting a CP. And the intonation might still be different in CPs functioning as incorporated pre-modifiers or lexicalized Ns. The analysis of these assumed differences is beyond the scope of the present study. 438 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS second syllable of any word or paradigmatic form. Words with first-syllable stress are explained by Chambers (1978) and Shaw (1980:30–53) as resulting from vowel epenthesis, vowel coalescence, or vowel deletion processes applied after the application of DSR or by the fact that the second syllable is a clitic. The words in (31) do not involve any of these processes and thus cannot be explained by the DSR. Furthermore, the forms in (31) cannot be accounted for by the Compound Accent Rule (CAR) formulated by Chambers (1978) and Shaw (1980:37–38), which states that in compound words the stress on the second member is reduced from primary to secondary. This is summarized by de Reuse (1994:204): [T]he DAR treats Lexical Compounds as one word, and there will be only one stress on the resulting form, which will be on the second syllable, regardless of whether this syllable is in the first or second element of the compound. In the case of Syntactic Compounds, every element of the compound is accented on the second syllable by the DAR, the compounding occurs, and then a further rule, called Compound Rule, weakens the accent that is on the second element of the compound. The DSR cannot explain the word-initial stress in the words given in (31), and these words do not conform to the definition of syntactic compounds as they have no secondary stress. The word ínuŋpa, in (31b), is listed by Shaw (1980:55) as one of the exceptions that the DSR cannot account for, and Shaw proposes that the stress position might be explained by a presence of a word boundary. Shaw presents the word as follows “/i#nuŋpa/ [ínùŋpa],” thus essentially stating that ínuŋpa has a secondary stress on the second member, which is not supported by the analysis of audio data. Stress-initial compounds with monosyllabic nominal components, like ípuzA ‘to be thirsty’, are in fact CPs consisting of <N1SV>, and their word-initial stress can be accounted for with the H* downstep in the following way: as the nominal co-predicate is monosyllabic and the SV co-predicate is stress initial, the two H* peaks are adjacent to each other, resulting in ‘tonal crowding’, which prevents the onset of the second H* peak. Thus the second H* peak is “undershot” to such an extent that it is essentially phonetically deleted. Under the DSR account the complete loss of the stress on the SV makes the construction look like a lexical compound with an unusual stress position, whereas in reality its prosodic properties originate from a syntactic process, in which the first co-predicate has a stronger H* peak relative to the H* peak of the second co-predicate. While constructions like those in (31) originate in complex predication, many of them undergo gradual lexicalization, which explains why the stress of the second member does not normally reappear when an affix is inserted before the second co-predicate, as in ímapuze ‘I am thirsty’. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 439 The prosody of syntactic compounds like ípuzA ‘to be thirsty’ is different from that of lexical compounds of the <N1SV> composition, whose first member is a monosyllabic N, in that the latter are always stressed on the second syllable. Additionally, lexical compounds are never Vs, but rather they are Ns that represent names of species, categories, classes, or geographical locations. This is illustrated with the contrastive data in (32), where (32a) is a CP, whereas (32b) is a lexical compound (it is a name of a plant). (32) (a) Čháŋ šašá ʔ. (data: IEC) AUDIO 32A čháŋ Ø-ša-šá wood INAN-red-REDUP They are red trees. (b) Čhaŋšáša ʔ. (data: NSB-9: 00:10) AUDIO 32B Ø-čhaŋšaša INAN-red.osier.dogwood It is red osier dogwood. Other examples of lexical compounds with monosyllabic Ns are heȟ áka (hé ‘horn’ 1 ȟ aká ‘branching’) ‘elk’, Ȟ esápa (ȟ é ‘mountain’ 1 sápA ‘black’) ‘Black Hills’, čhaŋzí (čháŋ ‘wood’ 1 zí ‘yellow’) ‘smooth sumac (Rhus glabra)’, and Mnilúzahe (mní ‘water’ 1 lúzahAŋ ‘rapid’) ‘Rapid Creek’. Whereas such names formed as lexical compounds are very common, we also find compounds in which the monosyllabic nominal member is stressed. Examples are hásapa (há ‘skin’ 1 sápA ‘black’) ‘a black-skinned person’, híŋzi (híŋ ‘fur’ 1 zí ‘yellow’) ‘a buckskin horse’, and híŋtȟ o (híŋ ‘fur’ 1 tȟ ó ‘blue’) ‘a roan horse’. As shown in this section, the word-initial stress of such compounds cannot be accounted for by the DSR but can be explained as resulting from a lexicalization of CPs. Thus, these are syntactic compounds, which do not conform to Chambers and Shaw’s definition. It will require more investigation to understand the motivation for choosing between the route of lexical and syntactic compounding. Tentatively it seems that most lexical compounds are names of entities (e.g., Ȟ eská ‘Rocky Mountains’) or of species (e.g., heȟ áka ‘elk’, čhaŋšáša ‘red osier dogwood’) whereas compounds resulting from the syntactic construction are classificatory descriptions of nominal categories, such as híŋtȟ o ‘a roan horse’. However, there are some exceptions that make such delimitation uncertain. Chambers’s definition of lexical compounds is problematic with respect to another type of compound: the incorporated pre-modification in which the first member keeps its stress, whereas the stress on the second member is reduced or lost if tonal crowding takes place. Consequently, if the first member is a monosyllabic word, the compound has a word-initial stress, as in thíkte ‘to murder sb’ 440 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS (literally ‘to house-kill’, from thí ‘house’ 1 kté ‘to kill sb’) and mnít’A ‘to drown’ (literally ‘to water-die’, from mní ‘water’ 1 t’Á ‘to die’). 9. Complex predicates as the nucleus of modifier phrases. This section addresses another outstanding issue of Lakota syntax that can be accounted for using the CP analysis. It is illustrated in (33), which shows two Ns in a clause headed by an intransitive verb. (33) Wičháša kiŋ pȟ éta wiyákpakpayela (data: BBBJ-35: 3:00, also yaŋkápe ló. in DT: 1-23) wičháša kiŋ pȟ éta wiyákpakpayela Ø-yaŋkÁ-pi yeló man the fire spark-ly/ling 3SG.U-sit-PL DECL.MSP The men sat with/having sparkly fires. (More literally: ‘The men sat fire-sparkly.’) AUDIO 33 Note that the words “with” and “having” are included only to enable a well‑formed English translation, but they have no counterpart in the Lakota sentence. The question here is as follows: since the RP wičháša kiŋ ‘the men’ is linked to the single argument of the intransitive predicate, what licenses the N pȟ éta ‘fire’ in this clause? There is no additional argument on the predicate that pȟ éta ‘fire’ could link to, and the N is not inside a PP either. In order to answer this question, we have to inquire about the syntactic function of wiyákpakpayela ‘sparkly (INAN. PL.)’. Traditional descriptions of Lakota grammar—for example, Riggs (1852), Buechel (1939, 1970), Boas and Deloria (1941), Rood and Taylor (1976, 1996), Ingham (2003), de Reuse (2006), and Ullrich (2008)—categorize words with the suffix ‑ya / ‑yela as adverbs and interpret their syntactic function as one of adverbials. This, however, is problematic, because if wiyákpakpayela in (33) were an adverbial, then it would modify the predicate and would have no syntactic relationship with the N pȟ éta ‘fire’. Conversely, if one were to claim that wiyákpakpayela modifies the N, then there would be no syntactic relationship between the modified N and the predicate. In reality, the sentence in (33) contains an <N1SV> sequence pȟ éta wiyákpakpa ‘they are sparkly fires’, and the suffix ‑yela attached to the SV has scope over both of its members because they are connected via nuclearcosubordination. The suffix ‑yela releases the <N1SV> sequence from its predicative function and allows it to constitute a modifier phrase (MP) that in turn modifies the predicate. Under this analysis neither pȟ éta ‘fire’ nor wiyákpakpayela ‘sparkling’ are left without a syntactic relationship, which is reflected in the constituent projection of (33) in Figure 11. It shows that ‑yela (a bound morpheme) is connected to the SV at the word level (NUCw level), and its scope over both members of the <N1SV> sequence is accounted for COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 441 FIG. 11—Modifier phrase functioning as an ad-CORE modifier with the NUC cosubordination at the phrase level. This is the same phenomenon seen earlier in (13) and (14), where the locative and the instrumental prefixes were shown respectively as having scope over both members of the <N1SV>. Note that the reduplication of wiyákpakpa ‘to be sparkly in each case (INAN. PL.)’ indicates the plurality of the fires, which is additional evidence for the syntactic relationship between the N and the ‑yela form. The suffix ‑yela liberates the <N1SV> from its predicating function and allows it to function as a modifier phrase. This MP is in the periphery of the core, and consequently it is a manner modifier in the broader sense (i.e., an adverbial phrase in traditional terms). The data in (33) shows that accounting for the composition and syntactic function of these very frequent modifier phrases is only possible through the nuclear cosubordination analysis of complex predicates composed of <N1SV>. It is useful to compare the construction in Figure 11 with that in Figure 12 (a projection of (34)), where the N pȟ éta ‘fire’ is modified with kiŋ ‘the’ and consequently functions as an syntactically optional RP linked to the argument of the predicate yaŋkÁ ‘to sit’. As a result, the N cannot form an NUC cosubordination with the modifier wiyákpayela ‘sparkling’, and consequently, the suffix ‑yela has scope solely over the SV, deriving it into a modifier. The modifier wiyákpayela ‘sparkling’ is an ad-ARG modifier rather than ad-core modifier, because if it were reduplicated, the reduplications would reference the participant and not the event (core). The ad-ARG modification is reflected in the constituent projection in Figure 12. Derived modifiers often have vague scope in Lakota, although many modifiers derived from SVs are semantically restricted to being participant oriented (ad-ARG or RP-external modifiers). 442 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS FIG. 12—Ad-ARG modifier (34) Pȟ éta kiŋ wiyákpayela yaŋké. (data: BBBJ-74: 23:00) AUDIO 34 pȟ éta kiŋ wiyákpayela Ø-yaŋkÁ fire the sparkle-MOD INAN-sit The fire was/sat sparkling. An important property of MPs is that they can be serialized (unlike SVs and CPs) or combined with simple modifiers, as shown in (35a) and (35b) respectively. (data: EDT-Bo: 227) (35) (a) Ištá ǧ iyéla pȟ ehíŋ žižíyela waúŋ na . . . ištá ǧ i-yé-la pȟ ehíŋ ži-ží-ye-la eye brown-DER-REST hair tawny-REDUP-DER-REST wa-úŋ na 1SG.A-exist and I live having brown eyes (and) tawny hair, and . . . (data: EDT-Col-3: 167) (b) Ištá sabyéla owóhiŋyaŋsyela awíčhayuta ké. ištá sab-yéla owóhiŋyaŋs-yela a-wíčha-Ø-yuta eyes black-DER cruel-DER look.at-3ANIM.U-3SG.A-stem ké HSY It was looking at them with its cruel black eyes. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 443 FIG. 13—(a) Serialized modifier phrase. (b) Serialized modifier (MP 1 DM). The constituent projections for (35a) and (35b) are shown in Figures 13a and Figure 13b respectively. In each of the structures, the serialized MP has scope over a core argument, because it attributes a property to the participant rather than expressing a manner of the event. Subject orientation is characteristic of MPs that involve body parts, and it emanates from the fact that body parts are possessed inalienably. The projections in Figures 13a and 13b are examples of syntactic constructions in which phrase-level constituents target word-level elements, such as the core arguments. This is not restricted to MPs involving body parts. Note for instance Wakȟ ályapi kiŋ kȟ alyéla Ø‑Ø‑yatké ‘He drank the coffee hot’, where the modifier kȟ alyéla ‘hot’ has scope over the object ARG, whereas the adcore interpretation is not possible because the following reading is rejected by native speakers: ‘He drank the coffee in a hot manner’. Modifiers derived from SVs via suffixation of ‑ya (‑yela, ‑yeȟ čiŋ, ‑yakel, ‑yelaȟ čiŋ) have traditionally been categorized as adverbs, but this categorization is not tenable, because these derived modifiers frequently function as ad-ARG modifiers and RP-external modifiers, which means that they are participant oriented (in addition to sometimes functioning as ad‑core modifiers). The participant versus event orientation of these modifiers is often restricted semantically, and at other times it is determined syntactically, whereas some modifiers are vague in their orientation. It should also be mentioned here that derived modifiers do not form a phonological compound with the predicate, as proposed by de Reuse (1994:201) following some of the transcriptions in Deloria’s texts and in Boas and Deloria (1941). Evidence against compounding is found not only in the analysis of audio data but also in syntax, because derived modifiers commonly occur in syntactic positions other than before the predicate.21 21 For a detailed discussion of the orientation and prosody of derived modifiers and modifier phrases, see Ullrich (2018). 444 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS Modifier serialization, exemplified in (35), is another strategy that allows ascribing multiple properties to a single N, as in (35b) / Figure 13b, or a single property to multiple Ns. The latter is achieved by using one and the same property word inside each MP in sentences like (35a) / Figure 13a, as, for instance in Ištá žižíyela pȟ ehíŋ žižíyela waúŋ ‘I am of tawny eyes and tawny hair’. This is in addition to those strategies discussed in 5. 10. Conclusion. This paper has presented a novel analysis of structures composed of adjacent noun and stative verb <N1SV> that constitute either post-nominal modification (when RP-internal) or complex predication (when clause-final). The complex predicate analysis makes it possible to account for numerous outstanding issues and previously unknown features of the language, including phenomena from phonology, prosody, morphology, syntax, and information structure. The syntactic analysis of complex predicates composed of <N1SV> shows that they are linked via nuclear cosubordination, which means that the two co-predicates are codependent and constitute a very tight syntactic construction in which no element can intervene between its two members. Nuclear cosubordination is a crucial theoretical tool to account for the fact that these CPs can be modified by certain prefixes and suffixes with scope over both co-predicates, which in turn makes it possible to explain the structure of complex transitive verbs, modifier phrases, and other constructions derived from CPs. Furthermore, the CP analysis resolves possessor coding in and the syntactic structure of <N1SV> constructions with inalienable nouns, which have not been convincingly accounted for in the extant literature. The prosodic analysis indicates that despite their syntactic tightness, complex predicates do not form a phonological compound, except for instances where the nominal component is monosyllabic and the SV component is stressed wordinitially, which results in tonal crowding that causes the loss of stress on the SV member. This, in turn, makes it possible to explain word-initial compounds for which previous studies, using the Dakota Stress Rule, could not account for. Consequently, the present study refines the definitions of lexical and syntactic compounding in Lakota, which cannot be based solely on stress position. In addition, this investigation showed that noun incorporation and compounding are much less common in Lakota than formerly claimed. The analysis of <N1SV> complex predicates (together with the analysis of modifiers derived from SVs and modifier phrases derived from CPs) represents a crucial innovation in the account of predication and modification in Lakota. Given the salience of <N1SV> complex predication in the morphosyntax of Lakota (and across the Dakotan subbranch), it can be predicated that it plays an important role in many, if not most, of the Siouan languages. COMPLEX PREDICATES WITH NOUNS AND STATIVE VERBS 445 Lakota is a language in which there is a lot of interaction between syntax and morphology. Syntactic rules commonly refer to morphological elements (as for instance modifiers that have scope over morphologically marked arguments). And conversely, morphological features often modify syntactic constructions (e.g., instrumental prefixes having scope over co-predicates connected via nuclear-cosubordination). This study shows that the Role and Reference Grammar can accommodate such intricate interaction between syntax and morphology. REFERENCES BOAS, FRANZ, AND ELLA DELORIA. 1941. Dakota Grammar. Memoirs of the National Academy of Sciences, vol. 23, second memoir. Washington, DC: US Govt. Printing Office. BUECHEL, E., S.J. 1939. A Grammar of Lakota. The Language of the Teton Sioux Indians. Rosebud, SD: Rosebud Educational Society. ———. 1970. A Dictionary of the Teton Dakota Sioux Language. Lakota-English: EnglishLakota, ed. by Paul Manhart, S.J. Pine Ridge, SD: Red Cloud Indian School Inc. ———. 1978. Lakota Tales and Texts. Pine Ridge, SD: Red Cloud Indian School. CARTER, R. T. 1974. Teton Dakota Phonology. University of Manitoba Anthropology Papers No. 10. CHAMBERS, J. K. 1978. Dakota accent. In Linguistic Studies of Native Canada, ed. E. D. Cook and J. D. Kaye, pp. 3–18. Vancouver: University of British Columbia Press. CHAMBERS, J. K. AND SHAW, P. A. 1980. Systematic obfuscation of morphology in Dakota. Linguistic Inquiry 11(2): 325–36. CHAFE, W. 1976. Givenness, definiteness, contrastiveness, subject, topic, and point of view. In Subject and Topic, ed. C. Li. New York: Academic Press. CUMBERLAND, LINDA A. 2005. A grammar of Assiniboine. PhD dissertation, Indiana University. DAHLSTROM, A. 1982. A functional analysis of switch-reference in Lakhota discourse. CLS 18: 72–81. DELORIA, E. C. 1932. Dakota texts. (American Ethnological Society, vol. 14.) New York: Stechert. DE REUSE, WILLEM J. 1978. A grammar of the Lakota noun phrase. MA thesis, University of Kansas. ———. 1994. Noun incorporation in Lakota (Siouan). International Journal of American Linguistics 60(3): 199–260. ———. 2006. Serial verbs in Lakota (Siouan). In Serial Verb Constructions: A Cross-Linguistic Typology, ed. Alexandra Y. Aikhenvald and R. M. W. Dixon, pp. 301–16. New York: Oxford University Press. GRACZYK, R. 2007. A Grammar of Crow. Lincoln: University of Nebraska Press. INGHAM, BRUCE. 2003. Lakota. (Languages of the World: Materials, vol. 426.) Munich: Lincom. JOHNSON, M., B. ROSEN, AND MATEJA SCHUCK. 2016. Evidence for a VP constituent in Hocąk. In Advances in the Study of Siouan Languages and Linguistics, ed. Catherine Rudin and Bryan J. Gordon, pp. 339–64. Berlin: Language Science Press. MERLAN, FRANCESCA. 1985. Split intransitivity: Functional oppositions in intransitive inflection. In Grammar inside and outside the Clause: Some Approaches to Theory from the Field, ed. Johanna Nichols and Anthony C. Woodbury, pp. 324–62. Cambridge: Cambridge University Press. MIRZAYAN, A. 2010. Lakota intonation and prosody. PhD dissertation, University of Colorado, Boulder. RANKIN, ROBERT L. 2002. A Diachronic Perspective on Active/Stative Alignment in Siouan. Ms., University of Kansas and Research Center for Linguistic Typology. RIGGS, S. R. 1852. Grammar and Dictionary of the Dakota Language. Smithsonian Contributions to Knowledge, vol. 4. Washington, DC: Smithsonian Institution. 446 INTERNATIONAL JOURNAL OF AMERICAN LINGUISTICS ———. 1893 [1977]. Dakota Grammar, Texts and Ethnography. Contributions to North American Ethnology, vol. 9. Washington, DC: Govt. Printing Office. ROOD, DAVID S., AND TAYLOR, ALLAN R. 1976. University of Colorado Lakhota Project: Beginning Lakhota. 2 vols. Boulder, CO: Department of Linguistics. ———. 1996. Sketch of Lakhota, a Siouan language. In Language: Handbook of North American Indians, vol. 17, ed. Ives Goddard, pp. 440–82. Washington, DC: Smithsonian Institution. ROSEN, BRYAN. 2016. On the structure and constituency of Hocąk resultatives. In Advances in the Study of Siouan Languages and Linguistics, ed. Catherine Rudin and Bryan J. Gordon, pp. 313–38. Berlin: Language Science Press. RUDIN, CATHERINE. 2016. Coordination and related constructions in Omaha-Ponca and in Siouan. In Advances in the Study of Siouan Languages and Linguistics, ed. Catherine Rudin and Bryan J. Gordon, pp. 369–95. Berlin: Language Science Press. SHAW, P. A. 1980. Theoretical Issues in Dakota Phonology and Morphology. New York: Garland. ———. 1985. Modularism and substantive constraints in Dakota lexical phonology. Phonology Yearbook 2:173–202. ULLRICH, JAN. 2008. New Lakota Dictionary, 1st ed. Bloomington, IN: Lakota Language Consortium. ———. 2016. Lakota Grammar Handbook. Bloomington, IN: Lakota Language Consortium. ———. 2018. Modification, secondary predication and multi-verb constructions in Lakota. PhD dissertation, Heinrich Heine University, Düsseldorf. VAN VALIN, ROBERT D., JR. 1985. Case marking and the structure of the Lakhota clause. In Grammar inside and outside the Clause: Some Approaches to Theory from the Field, ed. Johanna Nichols and Anthony C. Woodbury, pp. 363–413. Cambridge: Cambridge University Press. ———. 1987. Role of government in the grammar of head-marking languages. International Journal of American Linguistics 53(4): 371–97. ———. 2005. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge University Press. ———. 2008. Investigations of the Syntax-Semantics-Pragmatics Interface. Amsterdam: Benjamins. ———. 2013. Head-marking languages and linguistic theory. In Language Typology and Historical Contingency: In Honor of Johanna Nichols, ed. Balthasar Bickel, Lenore A. Grenoble, David A Peterson, and Alan Timberlake, pp. 91–123. Amsterdam: Benjamins. WEST, SHANNON L. 1999. The Pronominal Argument Hypothesis: The Status of Dakota Languages. Paper presented at Department of Linguistics Colloquium Series, University of Victoria. ———. 2003. Subjects and objects in Assiniboine Nakota. MA thesis, University of Victoria. WILLIAMSON, JANIS S. 1984. Studies in Lakhota grammar. PhD dissertation, University of California, San Diego.