Pragmatic Structures in Aymara

Petr Homola Matt Coler Codesign, s.r.o. INCAS3 Czech Republic The Netherlands [email protected] [email protected]

Abstract highly marked typological rarity [. . . ]. Still more idiosyncratic is the combination of the thematic The paper describes overt marking of in- marker with a grammeme combining the two func- formation structure in the indigenous An- tions of question and negation.” dean language Aymara. In this paper we We are not aware of any other language (except show that although the word order is free, for, to some extent, Quechua) in which word or- Aymara is not a discourse-configurational der is irrelevant to information structure. Hard- language (Kiss, 1994); instead, informa- man (2000) says about Aymaran languages that tion structure is expressed only morpho- “[t]he structural elements of a sentence may occur logically by pragmatic suffixes. The mark- in any order and are at the disposal of the speaker ing of ‘topic’ is more flexible that the for stylistic play.” Hardman et al. (2001) say that marking of ‘focus’ (be it at the clause level word order in Aymara “affects only style, not the or NP-internal). Since overt marking of grammar nor basic semantics.” Bossong (1989) information structure is partial, this pa- says about Quechua (which is typologically very per also devotes considerable attention to similar) that “word order is not only free but it the resolution of underspecification in Ay- is not primarily used as a means for expressing mara. pragmatic functions as such.” A detailed analysis of morphological information structure mark- 1 Introduction ing in Aymara can thus shed light on this module Chomsky’s original approach to formal syntax as- of grammar which is not expressed morphologi- sumes that sentences consist of constituents and cally in most languages. that the type and order of these constituents define Section 2 presents basic facts about Aymara. configurations that specify grammatical relations. Section 3 describes the overtly marked discourse As has been shown by Hale (1983), there are lan- categories in Aymara. Section 4 describes the guages in which word order has no or limited rele- means of morphosyntactic, semantic and prag- vance for grammatical relations and the respective matic referent identification. Section 5 treats prag- constituent trees are flat. In most such languages, matic marking in complex predicates. Finally, we word order is said to specify information struc- conclude in Section 6. ture (hence the name discourse-configurational languages coined by Kiss (1994), as opposed 2 Basic Facts about Aymara to syntax-configurational languages). We show Aymara is spoken by communities in a region en- that Aymara is neither syntax-configurational nor compassing Bolivia, Chile and Peru, extending discourse-configurational. In this language, in- north of Lake Titicaca to south of Lake Poopó, be- formation structure is expressed solely morpho- tween the western limit of the Pacific coast val- logically. This is very rare and therefore Ay- leys and eastward to the Yungas valleys. The lan- mara is very important for the research of informa- guage is spoken by roughly two million speakers, tion structure in general. As Bossong (2009) puts over half of whom are Bolivian. The rest reside it (referring to Aymara and Quechua): “Cross- in Peru and Chile. The Aymaran family (com- linguistically, grammemes explicitly expressing prised of Aymara, Jaqaru, and Kawki) is a lin- the function of theme are not very frequent; gram- guistic isolate with no close relative. Aymara is memes expressing the function of the rheme are a

98 Proceedings of the Second International Conference on Dependency Linguistics (DepLing 2013), pages 98–107, Prague, August 27–30, 2013. c 2013 Charles University in Prague, Matfyzpress, Prague, Czech Republic an affixal polysynthetic1 language (according to Due to overt morphological marking of infor- Mattissen’s (2006) classification) with a rich mor- mation structure, there are discontinuous phrases phology. It is SOV with modifier-head word order. in which the “gap” is not motivated by pragmat- Aside from the morphologically unmarked sub- ics, such as (2)2. The corresponding tree is given ject, all syntactic relations are case-marked, typi- in Figure 1. cally on the NP head. Roots can be divided into (2) Juma-n-x jiw-i-w nouns (including qualitative words), verbs, and you-gen-ref die-smpl3 3-npred particles. Suffixes, which may have a morpholog- → kimsa ch’iyar phisi-ma-xa ical or syntactic effect, can be classified as nomi- three black cat-poss2-ref nal, verbal, transpositional, independent, or prag- “Your three black cats died.” matic suffixes. The category of independent suffixes is comprised of three suffixes which are not jiwiw classifiable as members of either nominal or verbal NPRED morphology which likewise cannot be said to be pragmatic suffixes. These suffixes generally occur prior to inflectional morphology and/or the prag- phisimaxaREF matic suffixes. Pragmatic suffixes, by comparison, typically suffix to the last word of the entire sentence and/or the NP or VP. As their name suggests, jumanxREF kimsa ch’iyar their function is overall pragmatic in nature. Nom- inal and verbal suffixes can be subdivided into Figure 1: The surface syntax tree of (2) with a dis- inflectional and derivational categories, but given continuous noun phrase the ease with which category-changing transpositional suffixes attach to words of any category, of- Of course, the possessive pronoun jumanx could ten multiple times, it is common to find words with also form a continuous constituent with the rest of several nominal, verbal, transpositional, and inde- the NP. Or, it might be omitted since the posses- pendent suffixes. siveness is already expressed with the suffix -ma Detailed information can be found in (Hardman (phisimaxa). et al., 2001; Briggs, 1976; Adelaar, 2007; Cerrón- Overt marking of information structure is oblig- Palomino and Carvajal, 2009; Yapita and Van der atory in Aymara. A sentence without proper prag- Noordaa, 2008). matic suffixes (or with marking that would destroy intersentential cohesion) is considered ill-formed. 3 Overt Marking of Information It should also be noted that the marking of infor- Structure mation structure is orthogonal to the morphosyntactic expression of evidentiality.3 As mentioned in the introduction, word order in Aymara is not used to express information struc- 3.1 Referentiality ture. As an example, consider the following sen- Aymara has overt marking of referentiality. The tences in English (syntax-configurational), Rus- notion of referentiality roughly corresponds to sian (discourse-configurational) and Aymara (sub- 2 script N marks contextually new, i.e. nonpre- We use the following abbreviations in the glosses: ACC=accusative, AG=agentive, AGGR=aggregator, dictable information): ALL=allative, BEN=benefactive, CAUS=causative, COM=comitative, CONJ=conjectural, DIR=directive, (1) FUT=future tense, GEN=genitive, HON=honorific, HUM=human, IMPER=imperative, INCMPL=incompletive, (a) Peter cameN INF=infinitive, LIM=limitative, LOC=locative, MIR=mirative, It’s PeterN who came (*Came Peter) NEG=negative, NPRED=nonpredictable, PAST=past tense, (b) Pëtr prišëlN PART=past participle, POSS=possessive, POSS1=1st person Prišël Pëtr possessive, POSS2=2nd person possessive, POSS3=3rd N person possessive, POT=potential, PROGR=progressive, (c) Pedrox jutiwaN or JutiwN Pedroxa QM=question marker, REF=referential, REFL=reflexive, PedrowN jutixa or Jutix PedrowaN SMPL=simple tense, TRGRDS=transgressive (different subject), TRGRSS=transgressive (same subject), 1See (Baker, 1996). VRBL=verbalizer 3Aymara has a three-way system of evidentiality: wit- nessed, knowledge through language, and inferred. 99 what Sgall et al. (1986) call ‘contextual bound- he thought that A had left or was about to leave). edness’. We follow Sgall’s (2009) terminology Nonetheless, the verb is nonpredictable, too, be- here: “[Contextually bound] items are presented cause B does not know what exactly happened (A by the speaker as referring to entities assumed to may have come back earlier or may not have left be easily accessible by the hearer(s), in the pro- at all). totypical case ‘given’. They refer to ‘established’ Note that -xa marks referentiality, not items, i.e. to those that were mentioned in the pre- known/given information, as illustrated in (6). ceding co-text and thus still are sufficiently salient, (6) Qharüru-x sara-:-wa or to the permanently established ones (indexicals, tomorrow-ref go-fut1 3-npred those given by the relevant culture or technical do- → “I will go tomorrow.” main, etc.).”4 For the majority of bilingual speakers, referen- The adverbial expression qharürux “tomorrow” tiality is optionally marked with the suffix -xa,5 is marked as referential because it is inherently ref- however for monolinguals it is done so more con- erential but there is no marking which would spec- sistently. There are also slight dialectal differ- ify if it is predictable or not (the verb is already ences. marked as nonpredictable and there can be at most For example, in (3) the verb is marked as refer- one NPRED-marker per clause, as explained in the ential because the event is known from the context next subsection).8 (the speaker supposes that the hearer knows that It should be noted the the REF-marking is priva- someone came, otherwise he would not ask). tive, i.e. the absence of -xa does not mean that the (3) Khiti-s jut-i-xa? entity is not referential. who-qm come-SMPL3 3-REF → 3.2 Nonpredictability “Who did come?” Entities are nonpredictable if they are ‘new’ to On the other hand, in (5) the verb is not marked the hearer (cf. the usual dichotomy of informa- as referential because it expresses an event which tion structure known/given vs. unknown/new).9 is apparently unknown to the hearer (since he has Overt marking of nonpredictability in Aymara is expressed surprise in (4)). much more consequent than that of referentiality. In affirmative sentences, the most salient nonpre- (4) Aka-n-ka-sk-ta-sä dictable entity is marked with the suffix -wa.10 this-loc-vrbl-progr-smpl2 3-mir → The placement of the NPRED-marker can be best “Oh, you are here!” explained by question/answer pairs, such as (7) (5) Jïsa, wasüru-w and (8). yes yesterday-npred (7) Khiti-taki-s ut-x kutin-x-ta who-ben-qm house.acc-ref return-COMPL-SMPL1 3 → uta-ch-ta-xa “Yes, I came back yesterday.” house-caus-smpl2 3-ref → The adverbial expression wasüruw “yesterday” “For whom did you build the house?” is marked as nonpredictable6 (even though it is 8A reviewer suggested to compare pragmatic marking inherently referential) because it is more salient in Aymara with ‘topic’-markers in Japanese and/or Korean. than the verb7 (B is surprised to see A because First of all, whereas in the mentioned languages, topic marking interferes with case marking, in Aymara they are orthog- 4We do not use the terminology ‘topic-focus’ or ‘theme- onal. Also, in Aymara the use of pragmatic markers does not rheme’ because the paradigms of suffixes that express infor- affect word order. Finally, there are no free topics as in the mation structure do not correlate with it. For example, the following sentence from (Iwasaki, 2013): two paradigms are not always mutually exclusive (two suf- Zoo-waTOP hana-gaNOM nagai fixes from different paradigms can mark the same word). “The elephant; its nose is long.” 5In the literature this suffix is usually called “topic In Aymara, even if the word order of a NP is relaxed, all marker” or “attenuator”. parts of the NP agree with each other. 6This grammatical feature is described in the next subsec- 9Again, we follow Sgall’s (2009) terminology: “[Contex- tion. tually non-bound] items are presented as not directly pre- 7“More salient” means that it is more important in the cur- dictable in the given context, as ‘new’ information (at least rent context. Informally, we could also say “less predictable”. as chosen from a set of available alternatives).” The scale of salience is important for intersentential referent 10In the literature this suffix is often called “affirmative”, identification. “emphasizing”, or “declarative”. 100 (8) Jilata-ja-taki-w Noun phrases introduced by mä “one” are indef- brother-poss1-ben-npred inite and therefore always nonpredictable. Noun (uta-ch-t-xa) phrases with the determiners aka “this”, uka house-caus-smpl1 3-ref “that”, khaya (khä) “yonder” and khuri “away → “(I built it) for my brother.” yonder” are definite and therefore usually predictable unless they are explicitly NPRED-marked In negative sentences, the NPRED-marker is (in which case they are contrastive). The following usually attached to the negative particle jani, as example illustrates that referentiality and nonpre- in (9). dictability do not exclude each other.

(9) Jani-w kullaka-ma-r (15) Aka warmi-mpi-w mä jisk’a not-npred sister-poss2-all this woman-com-npred one small uñj-k-t-ti marka-n jiki-s-t-xa see-incmpl-smpl1 3-neg village-loc meet-refl-smpl1 3-ref → → “I did not see your sister.” “It is this woman whom I met in a small village.” However, if an argument of the verb is more salient than the verb (i.e., it is contrastive), it at- In verbal complexes, the main (nonﬁnite) verb tracts the NPRED-marker, as in (10). may be NPRED-marked, as in (16) and (17):

(10) Kullaka-ma-ru-w jan (16) T’ant’ ala-ñ-w sister-poss2-all-npred not bread.ACC buy-inf-npred uñj-k-t-ti mun-ta see-incmpl-smpl1 3-neg want-smpl1 3 → → “It is your sister whom I did not see.” “I want to buy bread.”

There are constructions in which NPRED-marks (17) T’ant’ al-iri-w are disallowed, for example in imperative con- bread.ACC buy-ag-npred structions or in those marked with the conjectural sara-sk-ta evidential sufﬁx: go-progr-smpl1 3 → (11) Chur-ita-ya “I am going to buy bread.” give-imper2 1-hon → 3.3 Deeply Embedded Nonpredictable “Give it to me, please.” Elements (12) Jut-chi-ni Nonpredictability is usually marked at clause level 11 come-conj-fut3 3 but it can occur inside a noun phrase if it → “Maybe he will come.” is required by the discourse context, as in (18) (from (Hardman et al., 2001)).12 NPRED-unmarked afﬁrmative sentences in the future tense tend to have imperative meaning: (18) Naya-n-x pusi-w I-gen-ref four-NPRED (13) Sara-ñäni uta-ja-x utj-itu go-fut4 3 house-poss1-ref exist-smpl3 1 → → “Let us go.” “As for houses, I have four.”

An affirmative sentence without NPRED- This example shows that REF and NPRED mark- marking in the future tense does not alter the ers need not appear at the end of a (semantic) context of the present discourse and may have phrase. Similarly, a nominal modifier may follow modal meaning (Hardman et al., 2001): 11Hajicováˇ et al. (1998) call this kind of information- (14) Nay sar-ä-xa structural marking proxy focus. 12In this example, pusiw is not contrastive, it is simply an I go-fut1 3-ref → answer to the question How many houses do you have? Of “I will go, right? / Should I go?” course, one could say just Pusiwa. 101 its governor if its information structure status dif- (24) Qullq-x allux-w fers from that of its head in which case both words money.acc-ref a-lot.acc-npred receive pragmatic marking, as in (19) (from (Hard- ap-kat-ta man et al., 2001)). put-dir-smpl1 3 → “I collected ALOT of money.” (19) Uka-x apilla-x that-ref oca-REF Thus Aymara, like many nonconfigurational luxu-cha-ta-wa languages, allows discontinuous constituents. For freeze-caus-part-npred example, in (22), the noun phrase sumaw chuq’ix “That is FROZEN oca.” “good potatoes” is discontinuous because aka markanx “in this village” is not part of its surface Aymara is a radically nonconfigurational lan- syntax subtree. guage.13 However, even languages from this class may pose word order constraints on “deeply” em- 3.4 Focalizers bedded phrases. Indeed, in Aymara noun phrases “Focalizers” (i.e. focusing operators, in English have more rigid word order than clause con- adverbs such as only, even, also, always, at least, 14 stituents. For example, while (20) is well- etc., see (Hajicováˇ et al., 1998)) are mostly ex- formed, (21) is ill-formed because the modifier pressed by suffixes in Aymara. In many cases, the suma “good” does not precede its head. word with a focalizing suffix attracts the NPRED- marking, as in (25)–(28). (20) Suma chuq’i-w aka good potato-npred this (25) Kimsa-ni-ki-w sar-i marka-n-x ach-u three-hum-lim-npred go-smpl3 3 → village-loc-ref grow-smpl3 3 → “Only three went.” “In this village grow good potatoes.” (26) Wawa-pa-x may-ni-ki-wa (21) *Chuq’i-w suma aka child-poss3-ref one-hum-lim-npred potato-npred good this “He has only one child.” marka-n-x ach-u village-loc-ref grow-smpl3 3 (27) Iki-ña-k-w → Intended meaning: “In this village grow sleep-inf-lim-npred good potatoes.” mun-t-xa want-smpl1 3-ref → Nevertheless the order of a nominal modifier “I only want to sleep.” and its head is not restricted if both have a pragmatic marker, as in (22)–(24). (28) Nay kuna-w sar-t-xa I what-npred go-smpl1 3-ref (22) Suma-w aka marka-n-x → “Even I went.” good-npred this village-loc-ref chuq’i-x ach-u Another pattern is attaching the independent ag- potato-ref grow-smpl3 3 gregator -sa to the focalized word together with → “GOOD potatoes grow in this village.” NPRED-marking of the verb, as in (29) and (30).

(23) Uta-cha-ña-taki-x qala-x (29) Naya-s sara-rak-t-wa house-caus-inf-ben-ref stone-ref I-aggr go-‘also’-smpl1 3-npred → alluxa-w apthapi-ña “I also went.” a-lot-npred gather-inf (30) Juma-ki-s yati-sma-wa “One must gather MANY stones to build a you-lim-aggr know-pot2 3-npred house.” → “At least you should know it.” 13In terms of the generative grammar, languages with the rule S + at clause level. → C 14Unlike, for example, in Latin, where the word order of The focalizing negative particle jani “not” at- NPs is less rigid although not completely free. tracts the NPRED-marking in unmarked cases (but 102 see (10)). Two focalizers can be combined, as in (33) Mä marka-n mä jisk’a imilla-w the case of jani and puni “always” which gives one village-loc one small girl-NPRED the form janipuni “never” or jani and raki “also” utj-i which gives the form janiraki “neither”. Both exist-smpl3 3 → forms attract the NPRED-marking. “There was a little girl in a village.” 3.5 Resolving Underspecification (34) Jupa-x Mariya suti-ni-wa As was demonstrated in the above description, she-ref María name-poss-npred overt marking of information structure in Aymara “Her name was María.” is partial, as it leaves some elements underspec- In (33), both noun phrases as well as the verb ified. Since Aymara is a radical pro-drop lan- are nonpredictable. The noun imilla “girl” is 15 guage (any argument of the verb may be unex- NPRED-marked (i.e., marked as the most salient pressed, not only the subject), predictable argu- part of the utterance) because in (34), it is referred ments are often omitted. As a consequence, direct to by the pronoun jupa “he/she”. objects are usually omitted if they can be inferred The analysis of utterances in texts such as sto- 16 from the context, as in (31). ries and narratives reveals that Aymara speakers consequently take intersentential cohesion into ac- (31) Tata-ja-taki-w count when they decide where to place pragmatic father-poss1-ben-npred markers. ala-sk-t-xa buy-progr-smpl1 3-ref → 4 Referent Identification “I am buying it for my father.” Surface and deep syntax as well as semantics oper- It follows that if a verb is NPRED-marked and ate on isolated sentences. Now we will discuss the has an overt object, the latter is nonpredictable, formalization of pragmatics, the level of discourse too, unless it is REF-marked, as in (32). context. For the purposes of this subsection we assume (32) Naya-w um-x that we have a discourse that consists of sentences I-npred water.acc-ref s1, . . . , sn and that we have the corresponding fea- wayu-ni-:-xa ture structures f1, . . . , fn. An entity we call a fea- carry-dir-fut1 3-ref ture structure that represents a person, an object or → “It is me who will bring the water.” an event (an event may be dynamic if described by a verb or statal if described by a nominal predi- The same holds for other verbal arguments. cate). Every entity has a special attribute, INDEX, The only ambiguity arises from inherently referen- to represent coreferences. tial expressions which appear REF-marked or un- The discourse context is formally a list of marked (e.g., qharürux/qharüru “tomorrow”) and indices (values of the INDEX attribute) C = this marking does not correlate with their nonpre- i , . . . , i . The sentences are processed one by h 1 mi dictability.17 one. At the beginning, C = . For every f , we do ∅ i the following: 3.6 Intersentential Cohesion 1. For every entity in f , we try to find its refer- Overt marking of information structure, described i ent in C (we describe below how referents are in the previous section, has crucial importance identified). If a referent was found for an en- for intersentential cohesion and can be helpful for tity, its index in C is moved to the beginning coreference resolution. Consider, for example, the of the list. Otherwise, a new index is assigned following two sentences: to the entity and prepended to the list. 15See (Cole, 1987). 16The sentence is an answer to the question Who are you 2. The index of the NPRED-marked entity is buying it for? moved (or prepended) to the beginning of C. 17It is possible that the category of referentiality is under- going a reanalysis as a result of language contact. A more There are various strategies of identification of detailed analysis of the speech of monolingual speakers is needed. referents in the preceding discourse that can com- bine to resolve ambiguities. 103 4.1 Morphosyntactic Referent Identification followed by Aymara has a rich system of switch-reference suf- (39) Jani-w usuri-:-ta-p fixes that help to identify referents in the discourse not-npred pregnant-vrbl-part-poss3 context. For example, the sentence yat-k-t-ti know-incmpl-smpl1 3-neg (35) Tumasi-xi → Thomas-ref “I did not now that shej was pregnant.” klasi-n-ka-ska-:n-wa the referent identification of the covert pronoun class-loc-vrbl-progr-past3 3-npref → which is the actor of usurïtap “that she was preg- “Thomasi was in the class room.” nant” is not morphosyntactically restricted (Ay- mara has one pronoun, jupa, for both “he” and may be followed by the following sentence: “she”) but it is semantically restricted. It is obvious that the semantic information of this kind has (36) Yatichiri-xj manta-n-isin-x teacher-ref enter-dir-trgrss-ref to come from the lexicon. Likewise, there will be lexically encoded semantic gender for words such nuw-i-wa as tayka “mother”, jilata “brother”, imilla “girl”, hit-smpl3 3-npred → etc. “When the teacherj entered, hej hit himi.” 4.3 Pragmatic Referent Identification The first sentence adds the index of the en- As described in Subsection 3.6, explicitly NPRED- tity Tumasixi “Thomas” to C. The second sen- marked entities are more salient than other non- tence adds the index of the entity Yatichirixj “the teacher” to C. Furthermore, there are two unre- predictable entities in the same sentence. If there solved (covert) pronouns that represent the actor is the sentence and the patient of nuwiwa “to hit”. In this case the (40) Tayka-ma-wi switch-reference transgressive suffix -sina speci- mother-poss2-npred fies that the actor of nuwiwa is the actor of man- yatichiri-ma-mpj jik-i-si tasinx “entered”, thus the actor of nuwiwa has the teacher-poss2-com meet-smpl3 3-refl index i. The patient is coindexed with the next → “Your motheri met your teacherj.” entity in C with which it agrees morphologically (e.g., in terms of animacy) or semantically. On the followed by other hand, the sentence (41) Usuta-:-ta-m-x

(37) Yatichiri-xj manta-n-ipan-x sick-vrbl-part-poss2-ref teacher-ref enter-dir-trgrds-ref yat-x-i-wa nuw-i-wa know-compl-smpl3 3-npred → hit-smpl3 3-npred “Shei already knew that you were sick.” → “When the teacher entered, he hit him .” j i j the actor of yatxiwa “s/he already knew” is coin- changes the indexes of the pronouns in the ma- dexed with taykamaw “your mother” because this trix sentence because the switch-reference trans- entity is NPRED-marked in the first sentence and gressive suffix -ipana specifies that the subject of therefore more salient (it precedes other entities in mantanipanx is different from that of nuwiwa. C). If we move the NPRED-marker in the first sentence to the “teacher”, the meaning of the second 4.2 Semantic Referent Identification sentence will change: If there is the sentence (42) Tayka-ma-xi mother-poss2-ref (38) Tumasi-mpi Marya-mpj Thomas-com Mary-com yatichiri-ma-mpi-wj teacher-poss2-com-npred uñj-t-wa jik-i-si see-smpl1 3-npred → meet-smpl3 3-refl “I saw Thomasi and Maryj.” → “Your motheri met your teacherj.” 104 (43) Usuta-:-ta-m-x Guaraní, in Aymara such constructions are formed sick-vrbl-part-poss2-ref syntactically (i.e. the complex predicate value is yat-x-i-wa not created in the lexicon). As for motion verbs, know-compl-smpl3 3-npred such as (45), there is a morphological alternative: → “He already knew that you were sick.” j (48) Jichha-x t’ant 5 Complex Predicates now-REF bread.ACC ala-ni-rapi-:ma-wa A special case of pragmatic marking represent buy-DIR-BEN-FUT1 2-NPRED → the so-called complex predicates, i.e. monoclausal “I will go to buy bread for you now.” predicates composed of (at least) two predicative elements. In such constructions, a (fully) semantic Further evidence that such predicates are mon- verb combines with a modal or auxiliary element oclausal is the fact that polypersonal agreement is to express complex predication such as causative, expressed on the auxiliary or modal verb, not the volitive, supine, etc. The concept of complex full one, as in (49) and (50). predicates, elaborated by Alsina (1996), has been applied to a number of phenomena and languages (49) Ch’uq alja-ñ-w including, for example, Hindi light verbs (Mo- potato.ACC sell-INF-NPRED hanan, 1994) or Turkish causatives (Çetinoglu˘ et mun-sma want-SMPL1 2 al., 2008). We will leave aside the rather compli- → cated formal treatment of these constructions in “I want to sell potatoes to you.” unification-based grammars (Homola and Coler, (50) Tump-iri-w jut-sma 2013) and focus on the linguistic description of visit-AG-NPRED come-SMPL1 2 these form in Aymara from the pragmatic point of → view here. “I came to visit you.” Sentences (16) and (17) are examples of com- It is noteworthy that a complex predicate with a plex predicates. The constructions contain only motion verb can have two complements, namely a one NPRED-marker, i.e. they are monoclausal and locative phrase and a verbal complement: should be represented by a single feature structure with a complex functor. For (16), we get: (51) Al-iri-w Chukiaw buy-AG-NPRED La.Paz.ACC (44) FUNC ‘muna-’ sara-:na  FUNC ‘ala-’  go-SMPL3 3 → ARGS subjact, ARGS subjact, pat     “He went to La Paz to buy it.”  * h +  dobjpat    i      Even more evidence for monoclausality can be For (17), we get: found in sentences like (52):

(45) FUNC ‘sara-’ (52) Chukiaw sara-ñ-w  FUNC ‘ala-’  La.Paz.ACC go-INF-NPRED ARGS subjact, ARGS subjact, pat   h   mun-t-x irnaqa-ña-taki-xa  * dobjpat +  i  want-SMPL1 3-REF work-INF-BEN-REF     →     “I want to go to La Paz for work.” In a more concise notation, (44) can be written as: In (52), the nominalized verb irnaqañatakixa “to work” depends on sarañw “to go” creating a (46) muna(subjact, ala(subjact, dobjpat)pat) long-distance dependency. Without considering sarañw muntx Likewise, (45) can be written as: the verbal complex a monoclausal predicate, it would be linguistically counterintu-

(47) sara(subjact, ala(subjact, dobjpat)pat) itive and computationally hard to parse the sentence. Unlike in languages with morphologically It is obvious from the mentioned examples formed volitive verbal complexes, such as that the NPRED-marker tends to attach to the full 105 verb. This fact supports the hypothesis that modal Mark C. Baker. 1996. The Polysynthesis Parameter. (deictic) syntactic elements should be considered Oxford University Press. synsemantic, i.e. at the level of deep syntax they Georg Bossong. 1989. Morphemic marking of topic should be represented as attributes of the nodes of and focus. In Michel Kefer and Johan van der Auw- their heads rather then autonomous nodes or fea- era, editors, Universals of Language, pages 27–51. ture structures. This approach, adopted by us, is Georg Bossong. 2009. Divergence, convergence, con- consistent with (Sgall et al., 1986). tact. In Kurt Braunmüller and Juliane House, edi- We omit from the discussion morphologi- tors, Convergence and divergence in language con- cally built complex predicates, such as causatives tact situations, pages 13–40. (e.g. (53)). In other languages where they are ex- Lucy Therina Briggs. 1976. Dialectal Variation in the pressed syntactically, they would be treated in the Aymara Language of Bolivia and Peru. Ph.D. thesis, same way as the constructions described above. University of Florida. (53) Jacha-y-t-wa R. Cerrón-Palomino and J. Carvajal Carvajal. 2009. Aimara. In M. Crevels and P. Muysken, editors, cry-CAUS-SMPL1 3-NPRED → Lenguas de Bolivia. Plural Editores, La Paz, Bolivia. “I made him/her cry.” Özlem Çetinoglu,˘ Miriam Butt, and Kemal Oflazer. (54) FUNC ‘caus’ 2008. Mono/bi-clausality of Turkish Causatives. In  FUNC ‘jacha-’  Proceedings of the 14th International Conference on ARGS subjact, pat ARGS dobjact Turkish Linguistics.  * " # +  h i    Peter Cole. 1987. Null Objects in Universal Grammar. 6 Conclusions Linguistic Inquiry, 18(4):597–612. We have described the overt marking of informa- Eva Hajicová,ˇ Barbara H. Partee, and Petr Sgall. 1998. tion structure and its pivotal role both in intersen- Topic-focus articulation, tripartite structures, and tential cohesion in Aymara as well as coreference semantic content. Kluwer Academic Publishers, resolution. Through an analysis of morphological Dordrecht. information structure marking with pragmatic suf- Kenneth L. Hale. 1983. Warlpiri and the grammar of fixes in this language, we illustrate the irrelevance non-configurational languages. Natural Language of word order for information structure. While & Linguistic Theory, 1:5–47. overt marking of referentiality is not always conse- Martha Hardman, Juana Vásquez, and Juan de Dios quent, nonpredictability is marked with strict regu- Yapita. 2001. Aymara. Compendio de estructura larity. Having demonstrated that nonpredictability fonológica y gramatical. Instituto de Lengua y Cul- can be marked both at clause level and inside a NP tura Aymara. and that focalizing suffixes tend to attract NPRED Martha Hardman. 2000. Jaqaru. LINCOM EUROPA. morphology, we explained how underspecification is resolved, noting, for example, that the overt ob- Petr Homola and Matt Coler. 2013. Causatives as Complex Predicates without the Restriction Opera- ject of a NPRED-marked verb is also unpredictable tor. In Miriam Butt and Tracy Holloway King, edi- unless it is REF-marked. We also treated the iden- tors, Proceedings of the LFG Conference. tification of morphosyntactic, semantic and pragmatic referents through a formalization of prag- Shoichi Iwasaki. 2013. Japanese. John Benjamins Publishing. matics. The specifics of pragmatic marking in complex predicates show how the NPRED-marker Katalin É. Kiss. 1994. Discourse Configurational Lan- typically attaches to the full verb thus supporting guages. Oxford University Press. the hypothesis that modal syntactic constructions Joanna Mattissen. 2006. On the Ontology and Di- should be considered monoclausal. achrony of Polysynthesis. In Dieter Wunderlich, editor, Advances in the theory of the lexicon, pages 287–354. Walter de Gruyter, Berlin. References Tara Mohanan. 1994. Argument Structure in Hindi. Willem Adelaar. 2007. The Languages of the Andes. CSLI Publications. Cambridge University Press. Petr Sgall, Eva Hajicová,ˇ and Jarmila Panevová. 1986. Alex Alsina. 1996. The Role of Argument Structure in The Meaning of the Sentence in Its Semantic and Grammar: Evidence from Romance. CSLI Publica- Pragmatic Aspects. D. Reider Publishing Company. tions. 106 Petr Sgall. 2009. Where to look for the fundamentals of language. Linguistica Pragensia, 19(1):1–35. Juan de Dios Yapita and J.T. Van der Noordaa. 2008. La dinámica aymara. Conjugación de verbos. Insti- tuto de Lengua y Cultura Aymara.

107