On Syntactic Integration and Semantico-Pragmatic Distribution of Mbya Guarani Purpose Coding Strategies

Abstract

The aim of this paper is to describe the available strategies for coding purpose relations in Mbya, focusing on their semantico-pragmatic distribution. According to Dooley, Mbya Guarani presents two main different strategies involved in the coding of purpose relations: aguã nominalizations and vy-dependent clauses, and among the latter a motion-cum-purpose (MCP) construction subtype is included. These strategies differ in regard to the semantic class of the main verb, referential continuity between the dependent and the main unit, and the fact that the dependent form can be negated independently from the main clause—thus, establishing different degrees of integration within the main unit in each case—, but overlap in same-subject contexts that involve a directed motion verb. However, according to our data, speakers do not use these constructions interchangeably, whereas aguã nominalizations portray an intended hypothetical outcome of the event or state-of-affair (hereafter SoA) expressed in the matrix clause; vy-dependent clauses and motion-cum-purpose constructions consistently trigger a result interpretation, entailing that the intended SoA was successfully accomplished.

1. Introduction

This paper offers a description of three different strategies employed to express purpose relations in Mbya Guarani, namely the vy-dependent clauses (1a), aguã nominalizations (1b), and a motion-cum-purpose (hereafter MCP) construction (1c).

(1)	a.	Jagua	o-o	ka’aguy-re	tatu	o-juka=vy
		dog	3.AC-go	woods-LOC	armadillo	3.AC-hunt=SS1
		‘The dog went into the woods in order to hunt an armadillo’
	b.	A-ñotỹ		avachi	ha-’u=aguã
		1SG.AC-plant		corn	1SG.AC-eat=NMLZ.FUT
		‘I plant corn to eat’
	c.	A-a		a-jau.
		1SG.AC-go		1SG.AC-take.bath
		‘I went to take a bath’

Our data shows that the MCP constructions (1c) and vy clauses (1a) reflect a closer semantic link between the SoAs involved, entailing the successful accomplishment of the SoA encoded in the purpose clause, and temporal and locational continuity between the SoAs encoded by the main clause and the dependent clause. On the other hand, aguã nominalizations (1b) exhibit a looser connection between the SoAs encoded by the main clause and dependent clause, as they allow location and temporal discontinuity, and the accomplishment of the purposive SoA remains in the hypothetical realm.

This paper is organized as follows: firstly, I will briefly introduce purpose relations from a cross-linguistic perspective (Section 2), followed by a summary of Mbya Guarani basic grammatical features (Section 3). I will then focus on describing each purpose construction (Section 4.1) according to three basic features: the semantic class of the main verb (Section 4.1.1), referential continuity (Section 4.1.2), and shared TAM and negation values between the main and the dependent clause (Section 4.1.3). Afterwards, in (Section 4.2), I will analyze Mbya purpose constructions in their specific discourse context, and through elicitation, following the guidelines proposed by (Matthewson 2004; Chelliah 2001), I will establish felicitous contexts of utterance for each construction. In (Section 5) I will focus on the finitization process in TG-dependent constructions which accounts for the balanced nature of these constructions in Cristofaro’s terms (Cristofaro 2003), as introduced in Section 2, and compare them to similar constructions in other languages of the family. Finally, in Section 6, I offer the conclusions.

2. Expressing Purpose Relations Cross-Linguistically

Every natural language has at least one strategy devoted to coding purposive relations between different events, where one SoA is presented as the desired hypothetical outcome of a previous SoA. Syntactically, although purpose clauses are usually included in the realm of adverbial clauses, their coding properties may either deviate from prototypical adverbial clauses (overlapping with strategies employed for coding complement clauses, relative clauses, or constructions included in the domain of deontic modality) or they may overlap with other adverbial clauses, mainly clauses of cause and reason (Thompson et al. 2007; Longacre 2007; Schmidtke-Bode 2009). Conceptually, purposes are inherently future-oriented in that the SoA encoded by the purpose clause invariably takes place after the SoA encoded by the matrix clause (Schmidtke-Bode 2009, p. 19). Furthermore, purpose clauses also have an intrinsic modal orientation, as they encode the speaker’s intentions of bringing about a certain SoA, encoded by the dependent unit. Therefore, TAM information can be easily left out in the dependent clause on economic grounds (Cristofaro 2003).

Regarding referential continuity, a preference for participant co-referentiality between the matrix clause and the dependent clause has been established as a cross-linguistically common feature, given that one participant’s desires are more likely to involve an SoA where that same participant has a relevant role, or they have more control over the bringing about of that SoA, or simply because they benefit from the outcome of the event (Cristofaro 2003; Schmidtke-Bode 2009). This is especially relevant for our analysis as the constructions under study overlap in same-subject contexts, which actually account for the vast majority of contexts.

Following Stassen (1985), Cristofaro (2003) classifies verb forms in purpose clauses as either balanced or deranked. Although balanced verb forms can occur in independent declarative clauses, deranked verb forms structurally divert from them, as they may lack certain categorial marking, such as TAM values or person agreement, or may exhibit specific morphology unlikely to appear in independent forms, such as case marking, adpositions, or nominalizers. In this sense, we find a cross-linguistic preference for deranked purpose clauses (102 languages in the World Atlas of Linguistic Structures sample) in disfavor of balanced purpose clauses (38 languages) (Cristofaro 2013). Nonetheless, 30 languages in the WALS sample also exhibit the co-occurrence of both deranked and balanced purpose constructions in the same language. Syntactically, two different purpose constructions in the same language may exhibit different degrees of finiteness or integration in the main unit, and, therefore, be reserved for specific events or participant sets, i.e., distinguishing the same subject from a different subject between the dependent clause and the main unit (Thompson et al. 2007). This scenario often results in specific communicative needs; thus, in languages that synchronically exhibit at least two different purpose constructions, each one is usually also associated with specific semantic or pragmatic contexts.

3. Mbya Guarani Basic Grammatical Features

Mbya Guarani is genetically included in the Guaranian languages sub-group of the Tupi–Guarani branch of the Tupian stock (Rodrigues 1958; Rodrigues and Cabral 2002), and is the most geographically widely spoken language of the Guarani group, having roughly 27,000 native speakers distributed in northern Argentina, southern Brazil, and Paraguay (Ladeira 2020). The data discussed in this paper were collected by the author during several field trips to the Ka’aguy Poty village (Aristóbulo del Valle, Misiones) from 2017 to 2022, which mostly included text collection. Elicitation sessions were also conducted remotely during 2020–2021 with a native Mbya speaker from the Perutí village (El Alcázar, Misiones). All secondary sources cited throughout this paper were produced by non-governmental organizations in collaboration with Mbya teachers in several bilingual schools located in the Misiones province. Secondary data from the Brazilian variety of Mbya, when included for comparative purposes, will be cited as such.

The orthography employed for the transcription of the primary data cited in this paper follows the conventions used by native speakers. Mbya has six oral vowels: a (written a), ɛ (e), i (i), o (o), u (u), ɨ (y), and their nasal counterparts (written: ã, ẽ, ĩ, õ, ũ, ỹ). The consonants are: p (written p), t (t), k (k), m (m or mb before oral vowels), n (n or nd before oral vowels), ŋ (g̃), g (g), β (v), r (r), ɲ (ñ), ʧ (ch), ʔ (‘), h (h), and dʒ (j).

3.1. Verbal Morphology

Mbya exhibits an active–stative split system which distinguishes, by means of different sets of person markers, A arguments in transitive verbs (2a) and S_A arguments in active intransitive verbs (2b), a- ‘1SG.AC’ in both examples, from P arguments in transitive verbs (2c), and S_P arguments in stative intransitive verbs (2c), che- ‘1SG.IN’. These forms are included in Set 1 and Set 2 (

Table 1. Mbya Guarani argument indexation system.

	Set 1 (AC)	Set 2 (IN)	Set 3 (1 > 2)
1SG	a-	che-
1PL.EXCL	ro-	ore-
1PL.INCL	ja-/ña-	ñande-/ñane-
2SG	re-	nde-/ne-	ro-
2PL	pe-	pende-/pene-	po-
3	o-	i-/ij-/iñ-/h-

), respectively.

(2)	a.	a-i-kytĩ	manji’o	b.	a-ña
		1SG.AC-3P-cut	manioc		1SG.AC-run
		‘I cut manioc’			‘I run’/’I ran’
	c.	Ndee	che-r-echa	d.	che-r-achẽ
		2SG	1SG.IN-R-see		1SG.IN-R-cry
		‘You see me’/’You saw me’			‘I cry’/’I cried’

In transitive clauses, the participant indexation system follows a 1 > 2 > 3 hierarchy, meaning that P arguments are cross-referenced by a Set 2 marker when a third person acts upon a second or first person or when a second person acts upon a first person, as can be observed above, cfr. i- in (2a) and che- in (2c). Special portmanteau forms from Set 3 (Table 1, below) are employed in contexts where the hierarchy is inverted, indicating when a first singular person acts upon a second person, the last being singular (3a) or plural (3b).

(3)	a.	ro-nupã	b.	a-po-nupã
		1 > 2SG-hit		1SG.AC-1 > 2PL-hit
		‘I hit you (SG)’		‘I hit you (PL)’

Regarding the expression of TAME information, Mbya Guarani counts on several particles employed for coding past tense, as the recent past marker kuri in (4a). Nonetheless, as can be observed in the data presented in (2), unmarked sentences in Mbya can receive both a present and past tense reading. Future tense, on the other hand, is obligatorily marked in all verbal stems (4b).

(4)	a.	re-karu	kuri	b.	re-karu=ta
		2SG.AC-eat	PST.REC		2SG.AC-eat=FUT
		‘You ate (recently)’			‘You will eat’

Other TAME values can be expressed by clitics that indicate, for instance, completive aspect (5a) deontic modality (5b), and evidential meanings, such as the reportative je (5c) and the indirect perception marker ra’e (5d).

(5)	a.	o-u=ma		b.	a-a=ra’ã	tetã-re
		3.AC-come=COMPL			1SG.AC-go=DEO	town-LOC
		‘He/she/they already came’			‘I have to go to town’
	c.	karai=je	o-u	d.	karai	o-u=ra’e
		man=REP	3.AC-come		man	3.AC-come=EVID
		‘They said the man came’			‘The man came’ [I heard him come in]

Finally, verbal negation is encoded by means of the circumfix n(d)-…-i (6a), whereas nominals can be negated by the constituent negation particle he’ỹ ~ e’ỹ (6b).

(6)	a.	n-a-karu-i	b.	a-echa	jagua-he’ỹ.
		NEG-1SG.AC-eat-NEG		1SG.AC-see	dog-NEG
		‘I do not eat’/’I did not eat’		‘I do not see a dog’/’I did not see a dog’ [I saw something else]

3.2. Simple Clauses

As described in the previous section, core arguments are obligatorily cross-referenced in Mbya Guarani verbs; thus, they are frequently lexically unexpressed: cfr. (7a), (7b), and (7c). However, in transitive clauses the language may exhibit SOV or SVO constituent order if both core arguments are lexically realized (i.e., first mention). Other orders such as OSV, although less frequent, are also possible when a certain constituent is fronted for pragmatic purposes.

(7)	a.	o-japo			V
		3.AC-make
		‘(he/she/they) made (it)’
	b.	o-japo		ajaka	VO
		3.AC-make		basket
		‘(he/she/they) made a basket’
	c.	Poty	o-japo		SV
		Poty	3.AC-make
		‘Poty made (it)’
	d.	Poty	ajaka	o-japo	SOV
		Poty	basket	3.AC-make
		‘Poty made a basket’
	e.	Poty	o-japo	ajaka	SVO
		Poty	3.AC-make	basket
		‘Poty made a basket’

3.3. Complex Sentences

Coordination in Mbya may be expressed by juxtaposition or by the conjunction ha’e, as in (8a), whereas subordination is marked by specific conjunctions or by switch-reference markers (9a and 9b, below). For instance, the subordinate conjunction jave in (8b) expresses temporal overlap.

(8)	a.	o-ky	ha’e	o-vera
		3.AC-rain	COOR	3.AC-shine
		‘It rained and there was lightning’
	b.	Che	pyregua	a-moĩta=jave
		1SG	shoe	1SG.AC-put.on=SUB.TEMP
		che-chu’u	ñandu.
		1SG.IN-bite	spider
		‘When I was putting on my shoes, a spider bit me’

According to the literature (Cabral and Rodrigues 2006; Dooley 1988; Dooley 2013), Mbya Guarani is the only language in the Guaranian group that developed a switch-reference (SR) system. Dooley (1988)2 describes the subordinating conjunction vy as a same-subject marker (SS), which occurs in dependent clauses whose subject is co-referential with the subject of the main clause, as opposed to ramo, different-subject marker (DS), which occurs when the subject of the subordinate and main clauses are not co-referential, as can be observed in the data below (cfr. 9a and 9b).

(9)	a.	[Ava	o-o	vy]	mbói	o-exa.
		man	3-go	SS	snake	3-see
		‘When the man was going, he saw a snake’					(Dooley 1988, p. 97)
	b.	[Ava	o-o	ramo]	mbói	o-exa.
		man	3-go	DS	snake	3-see
		‘When the man was going, the snake saw him’					(Dooley 1988, p. 97)

Mbya Guarani’s SR markers are compatible not only with temporal but also conditional and causal semantics, depending on the discourse context. Thus, a sentence such as (9a) may trigger a conditional reading ‘if the man goes, he sees a snake’, or a causal reading ‘because the man went, he saw a snake’ depending on the discursive context, and the same applies to ramo DS-dependent clauses.

4. Expressing Purpose Relations in Mbya: Basic Features

In this section, I will describe the different purpose coding strategies available in Mbya. Firstly, in Section 4.1 I will offer a synchronic analysis that focuses on three basic features in order to distinguish the degree of integration between the dependent form and the main unit in each construction. Then, in Section 4.2 I introduce an elicitation-based experiment developed in order to distinguish felicitous contexts of utterance for each construction.

4.1. Basic Features and Integration within the Main Unit

In the following sub-sections, I will focus on three basic features in order to distinguish the degree of integration between the dependent form and the main unit in the strategies employed for coding purpose relations in Mbya: the semantic class of the main verb (Section 4.1.1), the referential continuity between the main and dependent clauses (Section 4.1.2), and the sharing of TAM and negation values (Section 4.1.3). Finally, in Section 4.1.4, I offer some preliminary conclusions regarding the level of integration within the main unit in these three constructions.

4.1.1. Semantic Class of Main Verb

In regard to the semantic class of the verbs involved, these constructions exhibit different constraints regarding the main verb, i.e., the event that takes place in order to achieve the purpose encoded by the dependent form, whereas the dependent is an open-class verb. Motion predicates, for instance, are intimately related to purpose constructions cross-linguistically, as a purposive situation can easily be assimilated to one involving motion towards some target or place, this target or place being the realization of the dependent SoA (Cristofaro 2003, p. 176). Mbya motion-cum-purpose constructions consist of a basic purpose configuration, where the first verb of the series is a motion verb and the second an open-class verb, with no need of any dependency marking, as can be observed in (10a) and (10b).

(10)	a.	o-o	o-ma’ẽ.
		3.AC-go	3.AC-look
		‘He went to look’
	b.	e-ju	che-pytyvõ!
		2SG.IMP-come	1SG.IN-help
		‘Come help me!’

Only two directed motion verbs can participate in this construction in Mbya: ju ‘to come’ and o ‘to go’, cfr. (10) above. Motion verbs in these constructions do not exhibit a high degree of grammaticalization, since they actually indicate the direction of the motion event across space. Evidence for this is that they are often accompanied by locative FPs as kokue-py ‘to the field’ (11a) and ka’aguy-re ‘into the woods’ (11b).

(11)	a.	Kyrĩngue	o-o	kokue-py	o-eka		avachiky
		children	3.AC-go	field-LOC	3.AC-search		corn.cob
		‘The children went to the field to search for corn cobs’
	b.	Chee	a-a	ka’aguy-re	a-eka		ei
		1SG	1SG.AC-go	woods-LOC	1SG.AC-search		honey
		‘I went into the woods to search for honey’

Similarly, vy-headed dependent clauses are far more frequent with motion verbs as main predicates, but they can combine with more, as they also allow predicates such as o’ẽ ‘to go out’ (12a), and transitive motion predicates such as raa ‘to take’ (12b). Finally, though scarce in our data, vy purpose constructions can even allow other type of verbs as the main predicate, such is the case of juka ‘to kill’ in (12c).

(12)	a.	Ro-o’ẽ		oká-py		ro-ñevanga=vy.
		1PL.EXCL-go.out		house.exterior-LOC		1PL.EXCL-play=SS
		‘We went outside to play’
	b.	Che	memby	a-raa	tekoa-py	a-reko=vy.
		1SG	son	1SG.AC-take	village-LOC	1SG.AC-have=SS
		‘I took my son to the village in order to take care of him’ [lit: ‘in order to have him’]
	c.	Uru	ro-juka		ro-mo-ngaru=vy.
		chicken	1PL.EXCL-kill		1PL.EXCL-CAUS-eat= VY
		‘We killed the chicken in order to feed (them)’

On the other hand, aguã nominalizations show no restriction whatsoever with regard to the semantic class of the main verb, as they admit activities (13b), including motion predicates (13a), and, though rare in our data, even states, such as kuaa ‘to know’ in (13c).

(13)	a.	A-a	Tupã-reve	ha-’u=aguã	takuar-y	porã.
		1SG.AC-go	God-COM	1SG.AC-eat=NMLZ.FUT	takuara-water	good
		‘I go with God, in order to drink the takuara sacred water’ (Asociación Civil Coincidir—ACC 2018a, p. 35)
	b.	Ore	mbya-kuéry	ro-juka	ro-karu=aguã.
		1PL.EXCL	mbya-PL	1PL.EXCL-kill	1PL.EXCL-eat=NMLZ.FUT
		‘We the Mbya people kill in order to eat’
	c.	Kyrĩngue-i-ve-pe		a-ro-ayvu=aguã
		children-DIM-more-DAT		1SG.AC-COM.CAUS-talk=NMLZ.FUT
		a-i-kuaa-ve.
		1SG.AC-3P-know-more
		‘In order to talk with the more little ones, I am wiser’ [lit: ‘I know more’]

4.1.2. Referential Continuity

As stated in (Section 2), co-referentiality among participants in the main verb form and the dependent is a prototypical trait of purpose constructions. However, Mbya’s purpose coding strategies behave in different ways with regard to this feature. In Mbya motion-cum-purpose constructions, for instance, co-referentiality between the subject of the main verb form and the dependent is mandatory. Nonetheless, the S/A argument must be indexed in both verbs, such as in (14a) and (14b) where the active third person S/A index precedes the verbs o ‘to go’, tangara ‘to dance’ and pe’a ‘to peel off’.

(14)	a.	Ka’aru-kue	o-o	o-tangara	opy		r-oka-re
		1PL.INCL	3.AC-go	3.AC-dance	prayer.house		R-yard-LOC
		‘In the afternoon they go dancing in the opy yard’
	b.	Guavira	o-ĩ	ka’aguy-re,	kyrĩngue	r-ye	h-achy
		guavira	3.AC-be	woods-LOC	children	R-belly	3.IN-hurt
		jave,	ñande	chy	o-o	o-i-pe’a	i-pire.
		SUB.TEMP	1PL.INCL	mother	3.AC-go	3.AC-3-take.off	3.COR-skin
		‘the guavira fruit is in the woods, when children have a stomach ache, our mothers go to peel their skin off’ [lit: ‘to take off their skin’] (Asociación Civil Coincidir—ACC 2016, p. 23)

As can be observed in the preceding pair, Mbya motion-cum-purpose constructions admit both intransitive (14a) and transitive (14b) predicates, and transitive predicates in these constructions can introduce new arguments in discourse. In (14b), for example, the third person P argument i-pire ‘their skin’ is marked with the coreferential third person index, as it establishes a genitive relation with a referent in the previous discourse, namely, the guavira fruit. However, as can be observed in (15a) and (15b), these transitive predicates can also introduce new arguments, such as the verb eka ‘to look for’, which selects the O argument avachiky ‘corn cob’ (15a) or ei ‘honey’ (15b):

(15)	a.	Kyrĩngue	o-o	kokue-py	o-eka	avachiky
		children	3.AC-go	field-LOC	3.AC-search	corn.cob
		‘The children went to the field to search for corn cobs’
	b.	Chee	a-a	ka’aguy-re	a-eka	ei
		1SG	1SG.AC-go	woods-LOC	1SG.AC-search	honey
		‘I went into the woods to search for honey’

As for the vy-dependent clauses, as stated in (Section 3.3), vy serves as the same-subject marker in the Mbya switch-reference system, indicating that the subject of the dependent clause is co-referential with the subject of the main clause. In (16a), for example, both verbs are marked with the active first person index a-. However, in cases such as (16b) where the verbs are marked with a third person index, they are understood to predicate the same referent:

(16)	a.	A-ju	che	r-amoi	r-oo-py	a-iko=vy
		1SG.AC-come	1SG	R-grandfather	R-house-LOC	1SG.AC-live
		‘I came in order to live in my grandfather’s house’
	b.	o-o	ka’aguy-re	o-japo=vy	monde	haupéi	ñuã
		3.AC-go	woods-LOC	3.AC-make=SS	trap.type	and	trap.type
		‘He went into the woods to make armadillo traps and deer traps”

However, vy-dependent clauses can also admit same-subject sets (Dooley 1988), i.e., a plural subject that includes the subject referenced in the main clause. In (17), the dependent verb form is marked with a plural first person index ro- that includes the subject indexed by the first person singular index in the main verb a- and the adjunct che ryvy ‘my brother’.

(17)	che	r-yvy-reve	a-a	ka’aguy-re	[ñuã
	1SG	R-brother-with	1SG.AC-go	woods-LOC	trap
	ro-moĩ=vy		guachu	r-ape-py]
	1.PL.EXCL-put=SS		deer	R-path-LOC
	‘I went into the woods with my brother for us to put a trap in the deer’s path’
					(Asociación Civil Coincidir—ACC 2013, p. 35)

As for O arguments, as shown in (16b) and (17) above, the dependent verbs in vy-dependent clauses can introduce new arguments in discourse, such as monde and ñuã, different types of animal traps selected by the verbs japo ‘to make’ (16b) and moĩ ‘to put’ (17).

Nominalizations, on the other hand, do not exhibit any constraints regarding referential continuity, as can be observed in the following pair:

(18)	a.	a-vy	voi	a-mba’apo=aguã
		1SG.AC-wake.up	early	1SG.AC-work=NMLZ.FUT
		‘I woke up early to work’
	b.	A-mbopu	takuapu		kyrĩngue-’i	o-jeroky=aguã
		1SG.AC-play	takuapu		children-DIM	3.AC-dance=NMLZ.FUT
		‘I play the takuapu so that the children dance’   (Asociación Civil Coincidir—ACC 2016, p. 11)

If both the main and the nominalized verb are transitive, they can even select different O arguments, such as in (19a). However, it is relatively more frequent that the participants of the complex clause constitute a single participant pool, as in (19b) where the indirect object of the main verb and the subject of the nominalized form share the same referent chee ‘1SG’.

(19)	a.	O-gueru	jape’a	chee	a-mbojy=aguã
		3.AC-bring	firewood	1SG	1SG.AC-cook=NMLZ.FUT
		tembi’u-rã.
		food-FUT.NOM
		‘They bring firewood in order for me to cook food’
	b.	o-me’ẽ	che-vype	a-ñangareko=aguã
		3.AC-give	1SG-DAT	1SG.AC-take.care=NMLZ.FUT
		kavaju-re.
		horse-OBL
		‘He gave me the horse so that I take care of it’ (Associazione Cultural Rayuela—ACR n.d., p. 24)

Henceforth, although in vy-dependent clauses, if both the main and the dependent verb are marked with a third person index, they refer to the same subject, as in (16b) above, in aguã nominalizations they can refer to different subjects. In this case, the subject in the dependent is overtly expressed by an NP, in order to avoid ambiguities. In the following example (20), as both verbs, mbojera ‘to transform’ and upity ‘to get enough’, have active third person subject indexes, the subjects are overtly expressed by the NPs, tuu ‘the father’ and pavẽ ha’e javi-ete ‘each and everyone’, respectively.

(20)	T-uu		o-mbojera	ngu-ajy-pe		ka’a-i
	R-father		3.AC-transform	3.COR-daughter-DAT		yerba.mate-DIM
	pavẽ	ha’e	javi-ete	o-upity=aguã
	all	and	everyone-INTENS	3.AC-have.enough=NMLZ.FUT
	‘The father turned his daughter into yerba mate, so that it would be enough for all and everyone’ [lit: ‘so that everyone gets enough (of her)’]
					(Asociación Civil Coincidir—ACC 2018a, p. 24)

4.1.3. Sharing of TAM and Negation Values

As already stated in section (Section 2), purposes are intrinsically hypothetical and future-oriented, thus TAM marking is usually left out in purpose clauses. This is in fact reflected throughout my data. Mbya purpose constructions do not seem to exhibit specific TAM marking, as in most cases TAM values are solely marked in the main clause, as can be observed in the following examples (21) where the recent past marker kuri in (21a), the imperative mood prefix e- in (21b), and the indirect perception evidential ra’e in (21c) all occur in the main clause and affect the dependent verb form.

(21)	a.	O-u	kuri	ava	che-r-echa=aguã.
		3.AC-come	PST.REC	man	1SG.IN-R-see=NMLZ.FUT
		‘A man came to see me’
	b.	E-ju	che-pytyvõ!
		IMP-come	1SG.IN-help
		‘Come help me!’
	c.	O-o-py	ra’e	ka’aguy-re		o-eka=vy
		3.AC-go-DM	EVID	woods-LOC		3.AC-search=SS
		ngu-embi’u-rã-’i.
		3-food-FUT.NOM-DIM
		‘Turns out he went into the woods to search for food’

Negation, likewise, is traditionally considered a key feature to consider in defining the level of integration between the dependent and the main verb form in Mbya purpose constructions. In both aguã nominalizations and vy-dependent clauses, the purposive verb can be negated independently from the main predicate, as can be observed in (22a) and (22b), respectively. Mbya motion-cum-purpose constructions, on the other hand, do not exhibit a negative counterpart in our data.

(22)	a.	Re-o	tetã-re	nda-che-r-echa-i=aguã.
		2SG.AC-go	town-LOC	NEG-1SG.IN-R-see-NEG=NMLZ.FUT
		‘You went to town so you would not run into me’ [lit: ‘in order to not see me’]
	b.	A-a	nde-reve	nd-a-jeroky-i=vy.
		1SG.AC-go	2SG-with	NEG-1SG.AC-dance-NEG=SS
		‘I left so I would not have to dance with you’ [lit: ‘in order to not dance with you’]

It is worth noting that negation is constructed in vy clauses and aguã nominalizations by means of the circumfix used in main verbs, as opposed to the constituent negation marker he’ỹ ~ e’ỹ (6b), employed for other dependent forms, which also argues in favor of considering these structures as balanced, in Cristofaro’s terms (Cristofaro 2003, 2013). However, although the structures above are grammatical and fairly used by Mbya speakers, nowadays negative purpose is more frequently expressed by a specific avertive construction consisting of a clause introduced by the conjunction pono (23).

(23)	a.	Chee	a-moño’õ		ka’aru=jave,
		1SG	1SG.AC-gather		afternoon=SUB.TEMP
		pono	o-jeka=pa		upia.
		PURP.NEG	3.AC-break=TOT		egg
		‘I gather the eggs in the afternoon, lest they crack’
						(Associazione Cultural Rayuela—ACR n.d., p. 14)
	b.	Kya=ma		mitai	o-mo-ngue=aguã,
		hammock=COMPL		child	3.AC-CAUS-sleep=NMLZ.FUT
		pono	ho-’a	tupa-py	o-ke=vy.
		PURP.NEG	3.AC-fall	bed-LOC	3.AC-sleep=SS
		‘The hammock (is) to make the child sleep, lest they fall from the bed when they sleep’
						(Asociación Civil Coincidir—ACC 2018b, p. 16)

The conjunction pono is the reduced version of the Spanish loanword ‘para no’, employed in the source language for the same purpose, i.e., to introduce a possible unpleasant outcome one wishes to avoid by performing the event encoded in the main clause. This can be observed clearly in (23b): although aguã and vy occur in the pos-verbal position, pono, similar to its Spanish counterpart, occurs in clause-initial position. Pono clauses show no restrictions regarding the semantic class of the matrix verb or referential continuity among clauses, which is a cross-linguistically common trait of avertive or negative purpose constructions (Schmidtke-Bode 2009, pp. 129–44). Finally, it should be noted that, whereas language contact-induced loans often constitute lexical items, the borrowing of Spanish and Portuguese conjunctions is widely attested in languages of the TG group (cfr. Estigarribia 2020, p. 274 for Paraguayan Guarani and da Cruz et al. 2021 for Nheengatu) and other Native American languages (cfr. Guerrero 2017, p. 678 for Yaqui). However, the use of pono has not been documented for any geographical variety of Mbya (Dooley 2013; Martins 2004; Cadogan 1991).

4.1.4. Preliminary Conclusions on Integration between Units

In the previous sections, I have described Mbya purpose constructions, with regard to three main features: the semantic class of the main verb (Section 4.1.1), referential continuity (Section 4.1.2), and independent negation value (Section 4.1.3). A preliminary conclusion that can be drawn from this analysis is that these three constructions exhibit different restrictions. In this sense, referential continuity and independent negatability serve as useful parameters in order to distinguish the different degrees of integration exhibited by these constructions between the dependent verb form and the main clause (

Table 2. Integration within the main clause in Mbya purpose coding strategies.

	Semantic Class of Main Verb	Referential Continuity	Independent Negation
MCP	only general directed motion verbs: ‘come’ and ‘go’	only SS	no
vy	mostly motion verbs	SS and SS sets	yes
aguã	no restrictions	no restrictions	yes

), with the motion-cum-purpose constructions being the most integrated type, in opposition to aguã nominalizations, which appear to be the less integrated type of unit.

The different degrees of integration between the units are also reflected in the fact that aguã nominalizations can be fronted for pragmatic purposes, as can be observed in the example (29a) in the following section, whereas vy clauses and MCP constructions have to follow a sequential iconic order.

In the following section, semantic aspects of Mbya purpose constructions will be discussed, in order to distinguish which discourse contexts trigger the use of a specific construction by Mbya native speakers.

4.2. Semantico-Pragmatic Distribution

According to the analysis presented in section (Section 4.1), Mbya purpose coding strategies show different syntactic restrictions, but they still may overlap in specific scenarios. For instance, in same-subject contexts with a general directed motion predicate as the main event, such as in (24), all constructions previously described would be judged as being grammatical.

(24)		O-o-ra’ã	tetã-py	trámite-rami-ngua	o-japo=aguã.
		3.AC-go-DEO	town-LOC	paperwork-SIM-NMLZ	3.AC-make=NMLZ.FUT
		‘They had to go to town, in order to do a sort of paperwork’

However, a preliminary analysis of our data shows that, whereas aguã nominalizations portray an intended hypothetical outcome of the SoA expressed in the matrix clause, vy-dependent clauses consistently trigger a result interpretation, thus entailing that the intended SoA was perceived as successfully accomplished by the speaker. For example, the sentence in (24) was extracted from a narrative in past tense about an elderly couple that intended to go to town, but they actually were never able to do the paperwork they intended to, as there was a misunderstanding at the bus stop, and they ended up arriving there too late.

The data in this section were mostly obtained by elicitation, following linguistic fieldwork guidelines (Matthewson 2004; Chelliah 2001). In order to distinguish felicitous contexts for the various Mbya purpose constructions, I developed a simple two-step elicitation questionnaire. First, I presented the speakers with sentences taken from texts I collected in previous fieldwork trips or secondary sources which included a purpose relation, such as (25a), and then asked them if a different construction was felicitous in the same context.

		Sentences provided by the researcher:
(25)	a.	a-a-ju	a-ma’ẽ=vy	che	monde-re.
		1SG.AC-go-ITER	1SG.AC-look.at=SS	1SG	trap-OBL
		‘I went back again to look at my trap’
	b.	a-a-ju	a-ma’ẽ=aguã	che		monde-re.
		1SG.AC-go-ITER	1SG.AC-look.at=NMLZ.FUT	1SG		trap-OBL
		‘I went back again to look at my trap’
		Context provided by the researcher: ‘I placed a trap in order to hunt an armadillo and left it in the woods, came back the day after but no matter how hard I looked for it I could not find it. A man passed by and asked me why I was still in the woods.’

In the pair presented in (25), the only sentence judged to be felicitous by the speaker, given that context, was (25b), as (25a) would imply that the subject was actually able to look at his trap. Likewise, when presented with the pair below (26), only (26b) was deemed felicitous rather than (26a), as the men were not able to accomplish their desired purpose of working in the field.

		Sentences provided by the researcher:
(26)	a.	ro-o	ro-mba’apo=vy	kokue-py
		1PL.EXCL.AC-go	1PL.EXCL.AC-work=SS	field-LOC
		‘We went to the field in order to work’
	b.	ro-o	ro-mba’apo=aguã	kokue-py
		1PL.EXCL.AC-go	1PL.EXCL.AC-work=NMLZ.FUT	field-LOC
		‘We went to the field in order to work’
		Context provided by the researcher: ‘I went to the field with my son to harvest some corn cobs for dinner, but as soon as we got there it started raining so hard we had to rush back home, with no corn. My daughter saw us walk into the house dripping wet and asked me what had happened’.

Note that, although the accomplishment of the desired purpose is realized in vy clauses, aguã clauses seem to be rather undefined in that respect. In the following sentence (27), although the accomplishment of the event encoded by the vy clause is unmistakably understood as realized, that of the aguã clause is not strictly deemed as negative, but remains in the hypothetical realm. When asked for a context where the latter would be uttered, the speaker stated: ‘if the dog has not come back yet, maybe he has not hunted the armadillo just yet, or maybe I still do not know that, but I know that it was definitely his intention’3.

(27)		Jagua	o-o	tatu	o-juka=vy/=aguã
		dog	3.AC-go	armadillo	3.AC-kill=SS/=NMLZ.FUT
		‘The dog went to hunt an armadillo’ [context: the dog hunted an armadillo]

Likewise, in (28) the context provided a positive outcome of the event; however, both sentences are felicitous. Although the vy clause implies that the children were effectively cured by the old man’s medicine, a speaker may use an aguã clause in this context if, for instance, he lacks knowledge on the children’s well-being, or if the children just took the medicine and it takes time to have an effect.

(28)		Tuja-’i-va’e	o-japo	poã
		old-DIM-NMLZ	3.AC-make	medicine
		kirĩngue-pe	o-monguera=vy/=aguã
		children-DAT	3.AC-cure=SS/=NMLZ.FUT
		‘The old man made medicine in order to heal the children’ [context: the children were cured]

This distinction between accomplished and unaccomplished purpose is also reflected throughout our data in that vy purpose clauses are more frequent in past first person narratives, where the speaker usually knows how the events turn out and the accomplishment of the event encoded by the purpose clause is relevant for the narrative to go forward. On the other hand, aguã clauses prototypically occur in texts where the accomplishment of the event is less relevant, such as instructional texts, such as (29a), where a speaker is telling me how carved wooden figures are made, or in descriptive texts, such as (29b), where bees’ habits are described.

(29)	a.	Ja-japo=aguã		ta’anga,
		1PL.INCL.AC-make=NMLZ.FUT		figure
		Kurupi ka’y-gui	ja-japo=aguã,
		wood.sp-LOC	1PL.INCL.AC-make=NMLZ.FUT
		ja-a=ra’ã		ja-ru	ka’aguy-gui.
		1PL.INCL.AC-go-DEO		1PL.INCL.AC-bring	woods-LOC
		‘In order to make figures, in order to make the with kurupi ka’y, we have to go bring (it) from the woods’
	b.	Eiru	o-iko	yvoty’i-re,
		bee	3.AC-live	flower-DIM-LOC
		o-gueraa	ei-rã	ho-’u=aguã.
		3.AC-take	honey-FUT.NOM	3.AC-eat=NMLZ.FUT
		‘Bees live among flowers, (they) take nectar in order to eat’
					(Asociación Civil Coincidir—ACC 2013, p.22)

Motion-cum-purpose constructions, in this sense, behave in a similar manner as that of vy clauses as when in past tense, it is entailed that the purpose of the movement was successfully accomplished. This is to be expected because motion-cum-purpose constructions are the more syntactically integrated ones among Mbya purpose constructions. Nonetheless, these three strategies can also occur in present (30a) and future tense (30b). In these contexts, as the SoA encoded by the dependent verb form has not taken place at the moment of the utterance, the successful accomplishment of the event is out of question, but still speakers do not make use of these constructions interchangeably.

(30)	a.	E-ju	che-pytyvõ=vy/=aguã
		2SG.IMP-come	1SG.IN-help=SS/=NMLZ.FUT
		‘Come help me!’
	b.	A-a=ta	nde-r-oo-py	a-karu=vy/=aguã
		1SG.AC-go-FUT	2SG-R-house-LOC	1SG.AC-eat=SS/=NMLZ.FUT
		‘I will go eat at your place’

Although purpose constructions are intrinsically sequential, as the SoA encoded by the dependent form takes place after the SoA encoded by the main verb form, the time lapse between the two SoAs can vary. For instance, in cases such as (30a), the speakers make use of a vy-dependent clause if the event encoded by the purpose clause is located closer in time to the main event, an aguã nominalization, on the other hand, may entail the SoAs located further in time: “you could come now [afternoon] in order to help me tomorrow morning in the field” (see Note 3). Likewise, in (30b), a vy clause or a motion-cum-purpose construction would entail an immediate relation between the two SoAs ‘go’ and ‘eat’, whereas an aguã clause in the same context would lack this entailment. The same criterion applies to location continuity, as can be observed in the complex events portrayed by purposive chains in the data below (31).

(31)	a.	Kuñatai	o-o	kokue-py	o-gueru=vy
		women	3.AC-go	field-LOC	3.AC-bring=SS
		avachi,	o-japo=aguã		mbojape.
		corn	3.AC-make=NMLZ.FUT		bread
		‘The women went into the field to search for corn in order to make bread’
						(Associazione Cultural Rayuela—ACR n.d., p. 9)
	b.	Guembẽ	ij-aju=jave,		ro-o	ro-gueru
		plant.sp	3.IN-ripe-SUB.TEMP		1PL.EXCL.AC-go	1PL.EXCL.AC-bring
		ij-aju-va’e		ro-’u=aguã		ore-r-oo-py.
		3.IN-ripe-NMLZ		1PL.EXCL.AC-eat=NMLZ.FUT		1PL.EXCL-R-house-LOC
		‘When the guembẽ plant is ripe, we go bring those that are ripe in order to eat (them) at home’
						(Asociación Civil Coincidir—ACC 2016, p. 24)

In the pair above, motion-cum-purpose and vy clauses encode events located in the same spatial setting, whereas the event encoded by the aguã nominalization takes place in a different setting. This last fact is made explicit in (31b) as the purposive event encoded by the nominalization ‘in order to eat’ occurs with a locative adjunct oreroopy ‘at (our) home’.

The purpose of this section was to distinguish which discourse contexts are associated by native Mbya speakers with the use of each purpose coding strategy. As the data presented throughout this section show, vy clauses and motion-cum-purpose constructions entail a closer link between the events encoded by the main clause and the dependents than aguã nominalizations, in the sense that (i) they are usually associated with events located in closer spatial and temporal coordinates, and (ii) in past tense narratives, they entail the purpose of the event was successfully achieved. Both (i) and (ii) align with the cross-linguistic tendency of SR systems coding other types of continuity besides referential identity, such as location, temporal coordinates, or even modality (van Gijn and Hammond 2016).

5. Aguã and vy: Origins and Related Constructions

Regarding aguã, three genetically related proto-forms have been reconstructed for Proto-Tupi–Guarani (PTG) nominalizations4: (i) the action nominalizer suffix *-a ~ ø, e.g., * č-epják-a ‘my sleep’; (ii) the agent nominalizer suffix *-ár ~ cár ~ tár, e.g., *i-juka-cár ‘his killer’; and (iii) the circumstance nominalizer *-áb ~cáb ~táb, e.g., *i-juka-cáb ‘his death place or circumstance’ (Jensen 1998, pp. 539–41). Due to phonological changes in Mbya, as in other related languages (Rodrigues and Cabral 2002), this tripartite system has been simplified into a sole nominalizer -a, that serves all three functions. As can be inferred from the data reconstructed by Jensen, nominals derived by the prefixes presented in (i–iii) are cross-referenced by Set 2 person markers in PTG. Nonetheless, Mbya Guarani nowadays retains this indexation pattern only partially in agent nominalizations. Otherwise, nominalizations now follow the active–stative system similar to independent forms.

The nominalizer -a, with unmarked temporal reference, presents two counterparts with past and future connotations, angue and aguã, respectively. It is worth noting that aguã developed from the combination of the nominalizer and the anticipatory suffix *-rám ~ -ám ~ -wám, resulting in a meaning whereby the noun has not yet begun to serve its intended function (Jensen 1998, p. 510; Dooley 2013). Besides establishing purpose relations, as in (32a), aguã grammatical nominalizations cover several roles such as complements of perception and manipulative verbs, or agent and circumstance nominalizations with hypothetical future connotations. Hence, I restrain from labeling this morpheme strictly as a purposive marker or a complementizer (following Shibatani 2019; for an analogous analysis on Tapiete purpose clauses, see González 2018). In (32b), for example, the aguã nominalization serves as a complement to the verb jerure ‘to ask for’, and in (32c) it functions as noun with future connotations that, with the suffixing of -py ‘in’, serves as a locative.

(32)	a.	A-ñotỹ		[ha-’u=aguã].
		1SG.AC-plant		1SG.AC-eat-NMLZ.FUT
		‘I plant in order to eat’
	b.	Mbo’ea	o-jerure	kyrĩngue-pe	[o-estudia=aguã].
		teacher	3.AC-ask	children-DAT	3.AC-study=NMLZ.FUT
		‘The teacher asked the children to study’
	c.	O-o= ma		[o-jeroky=aguã]-py.
		3.AC-go= COMPL		3.AC-dance=NMLZ.FUT-LOC
		‘He already went to the party’ [lit: ‘to where they will dance’]

Finally, in Mbya, aguã behaves more like a clitic than a suffix, as it can follow other word-classes than verbs. For instance, in (33), aguã enclitisizes to voi ‘good’, which functions as an adverb modifying the verb guata ‘to walk’.

(33)	I-porã	poã	[kyrĩngue-pe	o-guata	voi=aguã]
	3.IN-good	medicine	children-DAT	3.AC-walk	early=NMLZ.FUT
	‘[It] is a good medicine for children to walk early’

In regard to vy, according to Rodrigues and Cabral 2003; Cabral and Rodrigues 2006, this morph also finds its origins in earlier nominalizations traced back to a stage before the separation of the Tupi–Guarani and Aweti families, involving the suffix *-áp ~ *-táp and the locative *–βo. The authors also stress that, in this earlier stage, purposive and temporal meanings were probably encoded by nominalizations combined with specific morphology, which were reanalysed as dependent clauses in many daughter languages of the TG group (this view is also shared by Schleicher 1998). As stated by Dooley (2013, p. 72), the reflex of this morph in Mbya gave rise to two different constructions: the vy-SS-dependent clauses that carry temporal, conditional and causal meanings, previously introduced in (Section 3.3), and the [V1 V2-vy] complex predicates5 (34a) which consist of an open-class verb and a movement or posture verb, that contributes aspectual, directional, or positional meaning to the construction.

(34)	a.	Ha’e	o-o-vy	o-purái.
		(s)he	3.AC-go-VY	3.AC-sing
		‘She went singing’
	b.	Ha’e	o-o	o-purái=vy.
		(s)he	3.AC-go	3.AC-sing=SS
		‘She went in order to sing’

Traditionally, the vy purpose construction (34b) has been described in other languages of the Guarani group as an instance of the vy complex predicate (34a), which would trigger a simultaneous interpretation when the suffix follows the movement verb or a purpose reading when the suffix is attached to the open-class verb (Cabral and Rodrigues 2006; Velázquez-Castillo 2004; Seki 2014; Jensen 1998). Nonetheless, as was already established throughout this paper and in a previous work (Vieira and Baranger 2021), our field data show that this construction expanded onto new syntactic domains in Mbya Guarani, and would be better analyzed nowadays as an instance of clausal subordination. Besides the S/A argument sharing restriction, the vy purpose construction does not display other structural constraints that the [V1 V2-vy] complex predicates do, as V1 belongs to a semantically open class, both verbs can be independently negated, V2 can select its own O/P argument independently of V1, and the morpheme under study behaves as a clitic more than as a suffix (Vieira and Baranger 2021, Section 6). This can be observed in (35), where V1 mbote ‘to close’ and V2 echa ‘to see’ select different O arguments, V2 is independently negated from V1, and vy is enclitisized after the last element of the negation circumfix.

(35)	A-mbote	che-r-echa	nd-oro-r-echa-i=vy.
	1SG.AC-close	1SG.IN-R-eye	NEG-1>2-R-see-NEG=SS
	‘I close my eyes in order not to see you’

As was observed throughout the data presented in this paper, the participant indexation system in Mbya Guarani dependent verb forms, as in nominalizations, follows the same active–stative system as that of main clauses. This system differs from the absolutive system reconstructed for PTG-dependent clauses, which codes P and S (both S_A and S_P) arguments with Set II forms (IN) (cfr. 36a and 36b), or uses special forms from the co-referential set in SS contexts6 (cfr. 37a and 37b).

(36)		Mbya Guarani
	a.	O-o	o-mbo’e=vy
		3.AC-go	3.AC-teach=SS
		‘He went to teach (someone)’
		PTG
	b.	*O-có	i-mo’é-bo
		3S-go	3P-teach-SER
		‘He went to teach him’		(Jensen 1998, p. 529)

(37)		Mbya Guarani
	a.	A-a	a-jeroky=vy
		1SG.AC-go	1SG.AC-dance=SS
		‘I went to dance’ PTG
	b.	*A-có	wi-poracéj-ta
		1S-go	1S.COR-dance-SER
		‘I went to dance’		(Jensen 1998, p. 530)

The fact that dependent clauses in Mbya, as in other daughter languages, nowadays display an internal finite morphosyntax similar to that of main clauses because they exhibit full-inflected predicates with finite features such as the active–stative indexation system found in independent clauses, circumfixal negation, and clause-level dependency markers, suggests they have undergone a gradual process of finitization, described as the extension of finite features7 to a non-finite construction without change in dependent status (Rose 2016, p. 350). Nonetheless, finite features still co-exist with other non-finite syntactic relics, such as unmarked verb forms when a lexically realized P argument immediately precedes the verb stem8, as can be observed in (38), analyzed by Schleicher (1998) as a nominalization involving a possessive construction, i.e., (38c) equals ‘I came for the boy’s meeting’ (Schleicher 1998, p. 217).

(38)		Mbya Guarani
	a.	A-a	monde-apo=vy
		1SG.AC-go	armadillo.trap-make=SS
		‘I went to make a trap’
		PTG
	b.	*O-úr	kunimí	∅-kuáp-a
		3-come	boy	LK-know-SER
		‘He came to meet the boy’			(Jensen 1998, as glossed in Magalhães 2021, p. 33)

The grammatical structure in (38a) is only exhibited by vy clauses in our data, and it is the only specific deranked feature exhibited by all three constructions, in terms that a verb stem only preceded by a noun with no personal indexes would be ungrammatical as the main predicate in a sentence9. Nonetheless, this specific configuration is not very productive nowadays, according to our data, as it only occurs with inanimate patients and verbs that prototypically select such patients, and the same verbs also allow for the more balanced configuration. Therefore, following Cristofaro (2003, p. 58), I propose to analyze vy-dependent clauses and aguã nominalizations as balanced forms, as the presence of an overt dependency marker or nominalizing morphology, which make these forms unable to occur as the main predicate of a sentence, is not enough evidence to consider them as deranked verb forms, and no other deranking strategies are at work.

In regard to Mbya Guarani MCP construction, I follow Dooley, who describes this construction as a subtype of the vy clauses, in which the dependency marker can be dropped when a directed motion verb serves as the main predicate (Dooley 2013, p. 76). In Cristofaro’s terms, this would also be considered a balanced form, because the purposive predicate can easily occur as the main predicate of a sentence, as the S/A argument is obligatorily indexed in both verbs, even though co-referentiality among participants is mandatory. Because of this last fact, in addition to the lack of overt dependency marking, two anonymous reviewers suggested Mbya Guarani MCP construction could be analyzed as an instance of a phenomenon cross-linguistically known as serial verb construction (SVC). Although, in general terms, I agree with this statement, I chose another term mainly because the definition of SVCs has been broadened so much since its origins to cover a wide variety of phenomena that I no longer find it transparent. Therefore, I opted for the term MCP construction in this paper, for the sake of clarity, as it is more specific and self-explanatory.

Nonetheless, it is worth noting that Rose (2009) identifies a re-ranked SVC in Emerillon, structurally similar to Mbya Guarani MCP construction10, that expresses motion serialization (39a) and sequential serialization (39b), among other meanings. Emerillon SVCs constitute an innovation that also encounters its origins in gerundive forms that gained finite features due to the finitization process (cfr. 34a in Mbya). Thus, their re-ranked nature resides in the fact that, as Mbya MCP constructions, they have lost the dependency marker and arguments are indexed in both verbs by means of the active–inactive indexation system, as can be observed in (39).

(39)	a.	Emerillon
		a-wedzu-tar		a-zaug
		1SG.I-go.down-FUT		1SG.I-bathe
		‘I am going down (to the river) to bathe’				(Rose 2009, p. 12)
	b.	o-poʔo	iŋa	i-dzupe	o-meʔeŋ
		3.I-pick	fruit.sp	3.II-to	3.I-give
		‘He picks an iŋa and gives it to him’				(Rose 2009, p. 26)

As stated by Dooley (2013, p. 76), the origin of Mbya Guarani MCP construction is probably similar to that of Emerillon SVCs, and, following Rose’s (2009) analysis, they would clearly be considered as such. In recent works, Magalhães (2021), on the other hand, analysed a genealogically related structure in Guaja, which also suffered the loss of the dependency marker and the shift in the indexation system, with sequential meaning (40a). Nonetheless, the author suggests it would be better analyzed as a paratactic sequence of independent predicates, as each verb can be independently negated, receive their own TAM marking and co-referentiality among participants is not mandatory (Magalhães 2021 p. 43).

(40)	a.	Guaja
		jaha	a-me’ ẽ		pyha	rahy	a-py’ỹ	kapo	kama	r-ia
		I	1SG.I-wake.up		night	INTS	1SG.I-get.out	POS	bed	LK-of
		a-kakaru		kakaruk-aha		Ø-pepe
		1SG.I-pee		pee-NMZR		LK-in
		‘I woke up late at night, got out of bed and peed in the bathroom’
	b.	jaha	a-me’ ẽ		pyha	rahy	a-py’ỹ	kapo	kama	r-ia
		I	1SG.I-wake.up		night	INTS	1SG.I-get.out	POS	bed	LK-of
		kakaruk-a		kakaruk-aha		Ø-pepe
		pee-GER		pee-NMZR		LK-in
		‘I woke up late at night and got out of bed to pee in the bathroom’
								(Magalhães 2021, p. 41)

As can be observed in (40b), in order to establish a purpose relation in Guaja with the same predicates of (40a), speakers use an adverbial clause headed by the dependent marker and the dependent predicate exhibits no personal indexes, a prototypical trait of Guaja intransitive dependent verbs (Magalhães 2021, p. 41). Therefore, although in Emerillon, SVCs with purposive and sequential meanings were both affected by the finitization process, in Guaja, only the sequential paratatic construction exhibits finite features, and was in fact more affected than its Emerillon counterpart.

The data from Guaja and Emerillon are interesting because, in order to express a sequential meaning with co-referential participants, Mbya Guarani speakers would also use a vy-headed dependent clause (41), as in purpose constructions. Nonetheless, the dependency marker cannot be dropped in this context, as it occurs in MCP constructions, which serves as an indicator of an earlier stage in the finitization process.

(41)	O-echa=vy	o-ñe-mondyi.
	3.AC-see=SS	3.AC-REFL-scare
	‘(She) saw (him) and got scared’

Emerillon, Guaja, and Mbya Guarani sequential and purposive constructions, therefore, show different stages of the same finitization process, which not only affects every daughter language to different degrees, but also affects different constructions in the same language to different degrees. It is also worth noticing that Emerillon-purposive SVCs co-exist synchronically with their less finite non-serial counterparts, as Mbya Guarani MCP constructions and vy-dependent clauses do, which is to be expected in an ongoing gradual finitization scenario (Rose 2016).

6. Conclusions

The aim of this paper was to describe in detail the current distribution of the strategies available in Mbya for the coding of purpose relations, namely motion-cum-purpose constructions, vy-dependent clauses, and aguã nominalizations. Firstly, according to the analysis presented in Section 4.1, the three strategies differ in terms of the semantic class of the main verb, referential continuity between the dependent and the main unit, and the fact that the dependent form can be negated independently from the main clause, establishing different degrees of integration within the main unit. Secondly, the analysis presented in Section 4.2 shows that the more integrated strategies, MCP and vy clauses, reflect a closer semantic link between the SoAs involved, entailing the successful accomplishment of the purposive SoA, and temporal and locational continuity between the main SoA and the purposive SoA. On the other hand, aguã nominalizations exhibit a looser connection between the SoAs, as they allow location and temporal discontinuity and leave the successful accomplishment of the purposive SoA unmarked. Finally, the finitization process that affected TG-dependent clauses accounts for the unexpected balanced nature of these constructions, as opposed to their PTG deranked counterparts, as shown in (Section 5).

The semantic distinction between the coding of purpose relations by vy clauses and aguã nominalizations align with what is cross-linguistically expected, as the SS marker vy is involved in the coding of other closely integrated relations at a cognitive level, such as cause and sequential relations, whereas aguã nominalizations are also employed for coding complements of manipulative, modal and desiderative verbs, all included in the hypothetical realm and intrinsically future-oriented (Schmidtke-Bode 2009; Cristofaro 2003), as was shown in (Section 5). In this respect, the borrowing of the avertive conjunction pono could be motivated by a drive for transparency (cfr. Guerrero 2017, p. 678 for parake/pake in Yaqui) given that both vy-dependent clauses and aguã nominalizations are employed to encode a wide range of semantic relations. Moreover, it is also cross-linguistically frequent that natural languages exhibit specific morphosyntactic strategies devoted to the expression of negative purpose that deviate from their positive counterparts (Cristofaro 2003, p. 158).