Guaraní Pronominal Arguments and Split Intransitivity

Julianne E. Hammink
Linguistics 504
May 6, 2010

Hammink 1
 1. Introduction
The South American language of Guaraní is one of the most widely spoken1 and described
indigenous languages of the Americas. Despite its ubiquitousness, Guaraní syntax is seldom
examined within a generative framework. In this paper, I will present some puzzling aspects of
Guaraní morphosyntax and demonstrate how theory independently proposed by Jelinek (1984)
for Warlpiri might also provide an explanation for the variation in Guaraní word order, as well as
for a split pattern of person agreement in Guaraní intransitive verbs. I will present these
phenomena in section 2, then describe Jelinek’s Pronominal Argument Hypothesis (PAH) in
section 3. Finally, in section 4 I will examine whether the predictions that the PAH makes for
Guaraní are borne out by the linguistic evidence.

2. Guaraní
Guaraní is a Tupian language2 spoken in Paraguay, Bolivia, Brazil and Argentina. In Paraguay,
Guaraní has been in contact with Spanish and Portuguese for almost 500 years. Typologically, it
can be considered a head-marking agglutinative language (Trinidad Sanabria, 1998) with
productive processes of both noun (Velazquez Castillo, 1996) and verb (Hammink, 2006)
incorporation. Guaraní employs postpositions to indicate indirect objects, but does not otherwise
overtly mark NPs for case. In addition, Guaraní demonstrates a variable constituent order and
variation in the argument marking pattern of verbs. These phenomena are relevant to the PAH,
and they will be described in the following sections.

2.1 Word Order

There is weak consensus on word order in Guaraní: Gregores and Suárez (1967) and Colijn
(2007) argue for a S V (IO) O order, but Gregores and Suárez note that such an order is a
preference, rather than a rule, and the corpus data from Colijn demonstrates that the position of
subjects, especially, can vary a great deal: in sentences with transitive verbs, the overt subject
appears with nearly equal frequency in pre- and post-verbal positions, while intransitive subjects
demonstrate only a slight preference for a preverbal position. The sentence in (1) is an example
of a transitive verb with a postverbal subject and an object that is only represented with a person
agreement marker on the verb, but not with an overt nominal:

1. xe-jopi peteî kava

1SG-sting one bee
‘A bee stung me.’ (Gregores & Suárez, 1967)

2.1.1 Subject/Object drop

Although subject nominals are very frequently missing from Guaraní sentences, as in (2), neither
subject nor object must be represented by an independent overt nominal in the sentence.

2. a-ha-se la-seyõra
1SG-go-want madam
‘I want to go, madam’ (Gregores & Suárez, 1967)

1
The Paraguayan variety is spoken by nearly 5 million people.
2
Sometimes described as a macrolanguage that subsumes several dialects spoken in different areas, including Mbyá,
Chiripá, Ñandeva, Aché, and Paraguayan Guaraní. (Grimes, et al., 1996)

Hammink 2
 3. a-prepará-ta yane-ve tempráno porã la-seyõra
1SG-prepare-FUT us-PP.for early well madam
‘I will prepare (food) for us very early, madam’ (Gregores & Suárez, 1967)

The sentence in (3) is glossed by Gregores and Suárez as transitive with the object ‘things’ which
in context refers to ‘food’, yet no overt morphology represents an object, either as an agreement
marker or as an independent nominal. This ‘object drop’ is common in Guaraní, and as Choi
(2000) observes, in Paraguayan Spanish as well, a phenomenon that she ascribes to prolonged
contact with Guaraní.

2.2 Split Intransitivity

Split intransitivity refers to a difference in the marking of the arguments of intransitive verbs in
some languages. vanValin describes one set of intransitive verbs as marking only a subject, and
another set as marking only a direct object. Split intransitivity considers the difference in
argument marking to have a semantic basis.

Intransitive verbs in Guaraní fall into two classes, which are distinguished by the class of person
agreement morphology that each verb type may take. Transitive verbs use the same two classes
of agreement morphology to mark either grammatical subjects or objects, according to a “person
hierarchy” frequently described in grammars of Guaraní. Mithun (1991) attributes the differences
in the Guarani split intransitive system to lexical aspect, or Aktionsart.

2.2.1 Person Agreement Morphology

Guaraní has two sets of verbal affixes that indicate agreement with one of the verb's arguments.
In descriptive grammars, these morphemes are considered to represent agreement with a verb’s
argument, rather than being the arguments themselves. The two classes of person agreement
markers are listed below:

Class Class
I II
1SG a/ai- xe-
2SG re- nde-
3SG o/oi- i/ij/iñ-
1PL
Incl. ja/ña3- ñande-

1PL ro- ore-

Excl.
2PL pe- pende-
3PL o/oi- i/ij/iñ-
Table 1: Guaraní verbal person agreement markers

3
When morphemes appear in pairs, the second is used in nasal environments.

Hammink 3
2.2.2 Subject agreement
Class I agreement prefixes can be used to indicate verb agreement with the subject of transitive
verbs, as in (x):

4. ai-pyhy- ta aguará -pe

1SG-catch-FUT vixen- PP.to
'I will catch the vixen' (Gregores & Suárez, 1967)

They can also indicate subject agreement in some intransitive verbs, as in (x):

5. jaguarete'i o-sê ij-yvykua -gui

tiger.DIM 3-exit 3POSS-burrow-PP.from
'The tiger came out from his burrow' (Gregores & Suárez, 1967)

2.2.3 Object agreement

In addition, Guaraní has a person hierarchy (x) that triggers object agreement in transitive verbs,
when the object is more prominent in the hierarchy than the subject, as in (x). Object agreement
is indicated by the use of an agreement marker from Class II, and there is no overt representation
of the 3rd person subject when object marking is triggered:

6. Person hierarchy: 1st Person > 2nd Person > 3rd Person4

7. xe- ʔu- ta- hĩna

1SG-eat-FUT-PROG
'He will be eating me' (Gregores & Suárez, 1967)

Class II agreement markers can also be used with a second class of intransitive verbs, as in (x):

8. xe- manduʔa
1SG- remember
'I remember' (Gregores & Suárez, 1967)

Class II agreement markers are also present in deadjectival and some denominal predicative
constructions, as in (x) and (x) respectively:

9. ij-akî- ma xe-ao amã- mê

3-wet-completely 1SG.POSS-clothing rain-PP.with
'My clothing is completely wet with the rain' (Gregores & Suárez, 1967)

4
Jelinek and Demers (1994; 704) describe a similar person hierarchy in the PA language of Straits Salish. They
describe the hierarchy in terms of local and nonlocal arguments, where local arguments are the 1st and 2nd person,
and nonlocal arguments are 3rd person.

Person Hierarchy: *nonlocal agents > local patients [*3 > 1,2]

Jelinek and Demers observe that sentences where a nonlocal agent is acting upon a local patient are
prohibited. Instead, a passive construction must be used.

Hammink 4
 10. (Xe) xe-kyse-ta
(I) 1SG-knife-FUT
'I will have a knife' (Gregores & Suárez, 1967)

The difference between intransitive verbs that take Class I person markers, like (x), and those
intransitives which take Class II person markers, like (x) and (x), has generally been ascribed to
semantics. In a later section of this paper a syntactic explanation will be proposed for the two
types of intransitive verb.

Finally, Class II agreement markers can also indicate possession in a DP, as in (x):

11. i-py ha i-po –gui

3POSS-foot AND 3POSS-hand -PP.from
'From his feet and his hands' (Gregores & Suárez, 1967)

3. Pronominal Argument Hypothesis

3.1 Jelinek’s Pronominal Argument Hypothesis

Jelinek (1984) observed that some properties of the language Warlpiri could be explained best by
considering the clitic pronouns/ agreement markers attached to the AUX to be the arguments of
the verb. Separate noun phrases, when present, are adjoined in non-argument positions, to
indicate topic or focus (Jelinek, 2006).

The fact that Warlpiri has two different Case-marking systems for independent nouns and for
clitics Jelinek takes as support for the PAH: she proposes that the true arguments of the verb, the
clitic pronouns, are case-marked as NOM or ACC in a system she calls G-case5, while adjoined
nominals are marked with non-argumental Case (for example, as the object of a preposition)that
follows an ERG/ABS pattern, and which Jelinek calls L-case. Jelinek adds that there may be
PAH languages which do not mark non-argumental nominals for case at all, but maintain the G-
case system for the clitic arguments.

Jelinek (2006) describes three diagnostics for PA languages: the absence of pro-drop, case-
marking differences between pronominal arguments and adjoined nominals, and the absence of
determiner quantification.

The absence of pro-drop is the result of the lack of agreement in PA languages. The pronominal
affixes on the verb are always unstressed, and, Jelinek argues, are topical: referring to old
information. Reflexives in PA languages are also always unstressed, as they are in pronominal
object positions, and therefore topical as well. Independent pronouns, when present, indicate
contrastive topic and focus and therefore can be stressed.

Case-marking differences arise from the different positions of pronominal arguments, in A-

positions, and DPs in A’-positions. As a result, PA languages mark pronominal arguments with
G-case, and DPs with L-case. Alternately, adjunct DPs may not be marked for case at all. If the

5
G-case and L-case are adapted from Hale (1983) in which G-case refers to ‘grammatical’ case that occurs with
arguments, and L-case to ‘lexical’ case that occurs with non-arguments.

Hammink 5
same Case-marking system can be used for both pronominal arguments and DPs, the the
language cannot be considered to be a pronominal argument language.

Finally, Jelinek (2006) observes that the PA languages she analyzes in that paper have adverbial
quantification, but no determiner quantification. DPs in PA languages have a default definite
interpretation and are not in A-positions, while determiner quantifiers are associated with
indefinite nouns in A-positions. Additional morphology is used in some PA languages to indicate
the indefiniteness of a noun; otherwise an indefinite interpretation is only possible in existential
sentences.

Jelinek (2006) mentions, but does not examine one other feature of PA languages which may be
of interest in an analysis of Guaraní. Head-movement- derived noun incorporation is impossible
in PA languages because, in such languages, DPs are never in the A-positions from which
movement is permitted to occur. Jelinek does recognize that in some PA languages a small class
of nouns exists that may appear in the verb complex as if they were pronominal affixes.

3.2 Hale
Like Jelinek, Hale (2003) argues that the only arguments of the verb in a PA language are the
pronominal affixes on the verb. Hale extended the PAH by arguing that the order of affixes on a
‘verb –word’ was exclusively the result of Merge, and that head movement was not necessary to
his version of the PAH. Hale observes that a non-movement account is more minimalist in spirit
in that it does not propose head movement that it would then need to motivate.

Hale also described the contrast between two types of intransitive verb: unaccusative and
unergative, with an elaborated VP structure. Hale (2003) and Hale & Keyser (1993, 2002)
describe the contrast between intransitive verbs as a difference in the location of the single
argument: if the argument is internal to the VP, then the verb is said to be unaccusative (or
ergative); if the argument is external to VP, in [Spec, vP], then the verb is said to be unergative.
Theta roles are entirely structurally determined in Hale & Keyser’s theory, and because argument
positions are limited, the inventory of possible theta roles is limited as well. The structure of the
unaccusative is shown in (12a) and the unergative in (12b):

12. a. v b. v

v V pronoun v

pronoun V v V

R V R V

In (12), the lexical verb consists of a root R that, in the case of (12a) projects a specifier in V,
while in the case of (12b), Hale argues that the root is nominal, and therefore cannot project a
specifier. The result is that the pronominal affix of an unaccusative verb is an argument internal
to the lexical V, while the pronominal affix on an unergative verb is external to it, in [Spec, vP].
The theta role of the argument in (12a) may be interpreted as ‘theme,’ and that of (12b) as
‘agent.’

Hammink 6
3.3 Predictions for Guaraní
Jelinek and Hale make specific predictions for the behavior of a PA language. In order to
establish that Guaraní is a PA language, it will be necessary to demonstrate the following
evidence of pronominal arguments: obligatory argument morphology on the verb, and optional
presence of independent DPs; phonologically null morphemes with fixed φ-features rather than
pro-type agreement; and adverbial, rather than determiner, quantification.

Because independent DPs are adjoined in A’-positions, their order relative to the verb complex
should be able to vary. Therefore, Guaraní should demonstrate a relatively free DP order if those
DPs are in fact adjuncts. In addition, the adjoined position of these DPs should prohibit noun
incorporation derived via head movement.

If a PA account of Guaraní can be established, it will serve to explain the contrast between Class
I intransitive verbs like (5) and Class II intransitives like (8) (as well as (9) and (10)). The
differences in these verb classes could be described as the contrast between the case marking of
the single arguments of unaccusative and unergative verbs, reflecting the position of those
arguments in the syntax.

4. Guaraní as a Pronominal Argument language

The PAH analysis can be applied to Guaraní if we consider the person-agreement markers listed
in Table 1 to be clitic pronouns which realize the arguments required by a verb. In sentences like
those in (5) and (8) above, such an analysis would be straightforward. The grammatical of the
intransitive verb represented by overt verbal morphology, and no other NPs are required. If
present, additional NPs would be in various adjoined positions, leading to the appearance of a
“free” argument order in such sentences.

Two problems remain for this analysis. First, transitive verbs are not always marked with overt
morphology that references both arguments, as may be observed in (3) and (7). Second, the PAH
as it stands does not explain why there should be two classes of intransitive verbs with different
systems for marking arguments.

In the following sections, I will present further evidence supporting a PAH analysis of Guaraní in
three areas: obligatory argument morphology on the verb in section 4.1, variable DP order in
section 4.2, and, finally, the unergative/unaccusative classes of Guaraní intransitive verbs in
section 4.3.

4.1 Additional Object Morphology

There are several other morphemes in Guaraní that indicate person relationships. There are two
portmanteau morphemes that indicate a first-person subject and a second-person object. There
are also morphemes that represent an unspecified object ‘someone’ or ‘something,’ and
morphemes for reflexive and reciprocal actions. These are summarized in Table 2, with
examples of their use following.

Hammink 7
Morpheme Gloss
ro- 1st person singular subject > 2nd singular object
po- 1st person plural subject > 2nd plural object
je/ñe- Reflexive / passive
jo/ño- Reciprocal
poro- Unspecified human plural object
mba’e Unspecified nonhuman plural object
Table 2: Other person agreement markers

The morphemes ro- and po- can be used in place of other person agreement morphology on
transitive verbs to indicate that a first person subject performs the action of the verb upon the 3rd
person object, as shown in (13) and (14):

13. ro- peté

1SG.2SG- slap
‘I slap you’

14. po- peté

1PL.2SG- slap
‘We slap you’ (Gregores & Suárez, 1967)

A reflexive action can be expressed with the morpheme je/ñe. The same morpheme in the same
position can also be interpreted as a passive construction. As a result, sentences like (15) can
receive either a reflexive or a passive reading:

15. a-je- jokua yvyra-rehe

1SG-RF- tie tree-PP
‘I tie myself to a tree’/ ‘I am tied to a tree’ (Gregores & Suárez, 1967)

Note that in (15), there is no NP external to the verb that refers to an actor. The same is true for
reciprocal sentences like (16)

16. ja-jo- pytyvo

1PLi-RC- help
‘We help each other’ (Gregores & Suárez, 1967)

An object can be marked on the verb even when it is a vague ‘someone’ or ‘something.’ The
morphemes poro and mba’e, in the same second position as the reflexive and reciprocal, express
that the object is undefined or general:

17. a-poro- pytyvo

1SG-UO.human- help
‘I help someone’ / ‘I help people (in general)’ (Ayala, 1993)

All the morphemes explained in (13) - (17) above occur on transitive verbs. While overt
nominals may co-occur in the Guaraní sentence, they are not required. This suggests that it is the

Hammink 8
person morphology on the verb which functions as the argument(s) of the sentence, rather than
any independent NP that may also be present. Jelinek (1984) observed a similar pattern in
Warlpiri, and proposed the ‘Pronominal Argument Hypothesis (PAH) as an explanation. In the
next section, I will review the PAH and discuss how it may be useful for explaining some of the
characteristics of Guaraní.

4.1.1 Object markers

In section 1, the various classes of Guaraní person markers were outlined, and it was
demonstrated that Guaraní has a robust system for indicating the arguments of a verb. In certain
conditions, both subject and object can be marked with verbal morphology, as can be seen in
section 1.3. When this occurs, either the subject and object are combined into a portmanteau
morpheme, as in (13) and (14), or else the reflexive, reciprocal or unspecified object marker
follows the subject marker, as in (15-17). The order of person agreement morphology relative to
the verb is as in (18), where (a) shows the order of independent subject and object morphemes,
and (b) shows the placement of portmanteau forms:

a. Subject Object 18.

Verb
b. Subject + Object

4.1.2 Noun incorporation6

What has not yet been established is whether this person marking morphology indicates
agreement with independent nominal argument external to the verb, or whether the person
markers are themselves the clitic pronoun arguments, as proposed in the PAH. One piece of
evidence that supports the second hypothesis is the possibility of object noun incorporation, as
described by Velazquez-Castillo (1996). Some bare nominals can appear in the second position
that is sometimes occupied by a reflexive, reciprocal, or unspecified object, as in (19a).
Generally, these nouns are body part terms, as in (19b), but non-body-part nouns can also appear
in this construction, as shown in (19c). The examples and glosses are from Velazquez-Castillo
(1996):

19. a. a- mba’e- jogua-ta ko-ka’aru

1SG- UO.thing- buy-FUT this-afternoon
‘I’ll go shopping this afternoon’

b. ai-po-peté la-mitã
1SG-hand-slap the-child
‘I slap the child in the hand’

c. a-vaka-ami-ta ko-pyhareve
1SG-cow-milk-FUT this-morning
‘I’ll do some milking this morning’

6
I use the term ‘incorporation’ without assuming, after Baker (1988), that the “incorporation” described here is the
result of head movement. Instead, I follow Hale (2003) who argues that the PAH is a non-movement system in
which the order of pronominal arguments is the result of external Merge.

Hammink 9
Velazquez-Castillo observes that a sentence like (19c) has an unincorporated alternative like
(20), in which it appears that the noun vaka ‘cow’ has not been incorporated by the verb.

20. A-ñami-ta pe-vaka moroti

1SG-milk-FUT that-cow white
‘I’ll milk that white cow.’

If the sentences in (19) represent transitive verbs with two pronominal arguments, they should
pattern with transitive verbs when combining with additional morphology that is sensitive to
transitivity. Velazquez-Castillo describes two such morphological tests of transitivity: the
causative morphemes mbo- and uka- and the nominalizer/reflexive/passivizer je/ñe.

4.1.3 Tests for transitivity

4.1.3.1 Causation as a test of transitivity
Guaraní has two causative verbal affixes, mbo- and -uka. The first is used with intransitive verbs,
and results in a transitive verb. The second is used with transitive verbs, including those that
have been transitivized by the first. Therefore, these affixes can be used to assess whether a verb
with an incorporated object, such as (19a) is treated as transitive or intransitive within the
grammar. Velazquez-Castillo reports that native-speaker acceptability judgments varied for
structures like (21a) and (21b), but that overall, native speakers did not fully accept either
structure.

21. a. ?? mba’e-jogua-uka ‘Cause (someone) to buy things’

b. ?? mo-mba’e-jogua ‘Cause (someone) to shop’ (Velazquez-Castillo, 1996)

This is problematic for a PAH analysis of Guaraní in that it does not clearly demonstrate that
sentences like (19 a-c) are transitive verbs with an incorporated object. It is possible that
speakers of Paraguayan Guaraní vary in their interpretation of such structures, with some
speakers treating a sequence like mba’e-jogua as a lexicalized intransitive verb, while others
view the same sequence as a transitive verb with an incorporated object. In section 4.1.4 I will
review some historical evidence that Guaraní used to have a very robust object-marking system
that has diminished over time. It may be that some Guaraní speakers are reanalyzing transitive
constructions as a result of the reduction of the object-marking system.

4.1.3.2 Nominalization as a test of transitivity

Velazquez-Castillo proposes another test for transitivity in the form of the affix je/ñe. She
describes this morpheme as a nominalizer that only combines with transitive verbs, but as can be
observed from the table and discussion in 1.3, this morpheme can also be analyzed as a reflexive
pronominal argument that serves as the object of a transitive verb. Velazquez-Castillo observes
that je/ñe cannot combine with verbs that, like (19a) have incorporated their object, as in (22):

22. *ñe- mba’e-jogua

NMN/RF UO.thing-buy
‘shopping’ (Velazquez-Castillo, 1996)

Hammink 10
Velazquez-Castillo argues that this construction is not permitted because the verb with its
incorporated object is underlyingly intransitive. However, if we assume, as I have suggested, that
je/ñe is a pronominal object, then the reasons for the ungrammaticality of (22) have a different
motivation: ñe and mba’e are both in an object position, so one of them is unlicensed.

4.1.4 Gaps in the paradigm

Modern dialects of Guaraní seem to lack a 2nd and 3rd person object agreement marker.
Accounting for these gaps in the object agreement paradigm would strengthen the analysis of
Guaraní as a PAH language in which all arguments of the verb are represented by pronominal
clitics.

4.1.4.1 Historical/comparative evidence for object argument markers

Jensen (1987) demonstrates that a 3rd person object prefix can be reconstructed for proto Tupi-
Guaraní (PTG). This object prefix *i- or y- appeared between the subject prefix and the verb.
Although this object prefix is not used productively in modern Guaraní, its position can be
observed in the so-called aireal verbs: transitive verbs whose person agreement prefix consists of
a Class I person marker followed by the phoneme /i/:

23. a. ai-kytĩ ‘I cut (something)’ (Canese & Decoud Larrosa, 1983)

b. oi-su’u ‘He/she/it bites (something)’ (Guasch, 1996)

Interestingly, Canese and Decoud Larrosa observe that when the person hierarchy triggers object
agreement, the phoneme /i/ disappears, as would be expected if both /i/ and the Class II person
marker indicate grammatical object:

24. (ha’e) ne-pytyvõ ‘(he/she/it) helps you’ (Canese & Decoud Larrosa, 1983)

Jensen demonstrates that some non-aireal verbs also bear evidence of the object agreement
marker. The modern Guaraní verb japo is the result of the reanalysis of the object marker as part
of the verb. In (25) the PTG sentence in (a) has become the modern Guaraní sentence in (b):

25. a. a-y-apo Reconstructed PTG

1SG-3-do/make

b. a-japo Modern Guaraní

1SG-do/make
‘I do/make (it)’ (Jensen, 1987)

4.1.5 A summary of pronominal argument marking in Guaraní

In the previous sections, it has been demonstrated that Guaraní has a number of processes
available for marking both subject and object arguments directly on the verb through the use of
person markers. These argument-marking patterns are summarized in Table (x) below:

Hammink 11
Subject Object Person Subject Marking Object Marking
Person
1 1 Class I je/jo reflexive/reciprocal
2 ro/po portmanteau ro/po portmanteau
3 Class I poro/mba'e unspecified object
noun incorporation
(PTG object prefix) Ø
2 1 Ø Person Hierarchy Class II
2 Class I je/jo reflexive/reciprocal
3 Class I poro/mba'e unspecified object
noun incorporation
(PTG object prefix) Ø
3 1 Ø Person Hierarchy Class II
2 Ø Person Hierarchy Class II
3 Class I je/jo reflexive/reciprocal
poro/mba'e unspecified object
noun incorporation
(PTG object prefix) Ø
Table 3: Guaraní Pronominal Arguments for transitive verbs

As the summary in Table 3 indicates, there is quite robust argument marking on transitive verbs.
In all but one of the cases where no overt argument marking occurs there is no ambiguity,
because the null form can only have one interpretation. Thus, a sentence like (3), repeated here
as (26), can only have one interpretation: that of a 1st person subject acting upon a 3rd person
object:

26. a-prepará-ta yane-ve tempráno porã la-seyõra

1SG-prepare-FUT us-PP early well madam
‘I will prepare (food) for us very early, madam’ (Gregores & Suárez, 1967)

In the case of sentences where the person hierarchy is triggered, there is only one situation in
which a possible ambiguity may occur. When Class II morphology is used to indicate a 1st
person object, the person of the subject may be ambiguous.

27. xe-su’u
1SG-bite
‘You/he bites me’

In Guaraní, disambiguation may optionally occur through the use of an independent pronoun or
nominal, as in (1), repeated here as (28):

28. xe-jopi peteî kava

1SG-sting one bee
‘A bee stung me.’ (Gregores & Suárez, 1967)

Hammink 12
 As observed above, the position of these independent nominals is quite variable, and this
variability can be explained by the PAH, if such phrases are considered to be adjuncts, rather
than arguments. In the next section, this possibility will be explored.

4.1.6 Definiteness and adverbial quantification

Jelinek (2006) describes the distinction between determiner quantification and adverbial
quantification with the following examples from English:

29. a. determiner quantification: most children, every child, three children, few children
b. adverbial quantification: always work, never work, work often, just work
(Jelinek, 2006)

Jelinek claims that determiner quantification, as in (29a) should be impossible in PA languages,

because such quantification is associated with A-positions, and DPs in PA languages do not
appear in such positions. Adverbial quantification is abundant in Guaraní, as in (30):

30. o-ñe’ê ava-ñe’ê-mê mante

3-speak man-speech-PP.in ADV.only
They only speak Guaraní. (Gregores & Suárez, 1967)

Determiner quantification, as in (29a), is expressed predicatively in Guaraní, as in (31):

31. Heta xe-ryma-kuera a-reko

Be.many my-animals-PL 1SG-have
I have many animals. (Gregores & Suárez, 1967)

Contrary to Jelinek’s prediction for PA languages, Guaraní does allow DPs to contain number
quantifiers. In some cases the number peteĩ ‘one’ is used as a marker for indefiniteness, as in
(32), while in other cases the number may in fact be a quantifier, as in (33):

32. o-henú peteĩ ayvú

3-hear one noise
He heard a noise (Gregores & Suárez, 1967)

33. Mokõi tupasỹ ha peteĩ santo

Two god.mother CONJ.and one saint
Two virgins and a saint.

This is problematic for a PA analysis of Guaraní. Vieira (1995), in an analysis of the Tupi-
Guarani language of Asurini do Trocará, claimed that despite the evidence in sentences parallel
in structure to (33), that numbers like mokõi ‘two’ are not in determiner positions in that
language, and that all quantification in Asurini is adverbial.

Hammink 13
4.2 An explanation for word order variation
The Pronominal Argument Hypothesis was first proposed by Jelinek (1984) as an explanation of
the apparent non-configurationality of languages like Warlpiri. By re-analyzing pronominal
verbal affixes as verbal arguments and DPs as adjuncts, Jelinek accounted for the variable word
order of such languages.

Canese and Decoud Larrosa (1983) provide the following examples of synonymous Guaraní
sentences with dramatically different word orders:

34. a. kuehe o- Ø -japo oréve-ĝuarã xe-sy chipa

yesterday 3-3-make us-PP.for my-mother bread

b. chipa o- Ø -japo kuehe xe-sy oréve-ĝuarã

bread 3-3-make yesterday my-mother us-PP.for

c. o- Ø -japo oréve-ĝuarã kuehe chipa xe-sy

3-3-make us-PP.for yesterday bread my-mother

d. xe-sy o- Ø -japo chipa oréve-ĝuarã kuehe

my-mother 3-3-make bread us-PP.for yesterday

‘Yesterday, my mother made bread for us’

Under the PAH, the independent nominals xe-sy, chipa and oréve-ĝuarã are in non-argument
positions, while the pronominal arguments co-referent with xe-sy ‘my mother’ and chipa ‘bread’
are marked on the verb as ‘o-‘ and Ø, respectively. The NOM and ACC case of these
pronominal arguments is assigned configurationally, according to Jelinek (1984) and Hale
(2003), and in the following section I will examine how the split intransitivity pattern of Guaraní
intransitive verbs can be described configurationally.

4.3 Split intransitivity

The PAH, especially in the form proposed by Hale (2003) offers an explanation for the so-called
split intransitive pattern in Guaraní intransitive verbs. Recall that intransitive verbs in Guaraní
take a person marker from either Class I or Class II, and that, under the PAH, these person
markers are the pronominal arguments of the verb. Both classes of these pronominal arguments
are also used with transitive verbs: Class I when the agentive subject is overtly represented, and
Class II when the theme object is overtly represented, in a phenomenon that has been frequently
called ‘object agreement.’ The fact that the pronominal arguments of some Guaraní intransitive
verbs resemble transitive subjects, or ‘agents,’ while others resemble transitive objects or
‘themes’ suggests that these two types of intransitive argument occupy different positions in the
syntax. If we adopt the structures proposed by Hale (2003) shown in (12), then Guaraní
structures for both classes of intransitive verbs would be as in (35), where (35a) corresponds to
Class II verbs like (8) and (35b) to Class I verbs like (5):

Hammink 14
 35. a. v b. v

v V pronoun v

pronoun V’ v V

R V R V

Ø xe- manduʔa Ø o- Ø sê- Ø

v 3- memory V 3- v exit V
‘I remember’ ‘He/she/it goes out’

In these structures, Hale decomposes the VP into a lexical root R, which is a lexical item of
indeterminate grammatical category which merges with a null verbal head V. If the R is
predicative, then the V’ will project a specifier position, as seen in (35a). If the R is not
predicative, then no specifier position will be projected, as in (35b). These two structures capture
the distinction between intransitive Guaraní verbs that has been described in semantic terms as
active/stative by Klimov (1974) and Mithun (1991). Verbs which appear to mark their subjects
as agentive are those with an external argument, as in (35b), while those which have been
described as marking their subject as a stative experiencer are those in which the argument is
VP-internal.

A potential problem for this account of intransitive Guaraní verbs lies in the causative
construction. Guaraní has a causative head mbo-7 which functions as a light verb that introduces
an additional argument, resulting in the transitivization of intransitive verbs.

36. a. i-vai xe yvotytý

3-ugly 1SG.POSS garden
‘My garden is ugly’

b. yvytu o-mbo-vaí xe yvotytý

Wind 3-CAUS-ugly 1SG.POSS garden
‘The wind made my garden ugly’

37. a. o-pupu la-y

3-boil water
‘The water boils’

b. Rodolfo o-mbo-pupu la-y

Rodolfo 3-CAUS-boil water
‘Rodolfo boils water’

Hale (2003) argues that such transitivity alternations should only be possible with unaccusative
Class II intransitives like (36) because the VP internal argument of the intransitive verb becomes
the theme object of the verb complex, when transitivized with mbo-. Because the argument of
7
mo- in nasal environments.

Hammink 15
(37) is in [Spec, vP], it will not be a part of the verb complex if the verb is transitivized with
mbo- . Hale claims that the position of the argument is what prevents unergatives like (35b) and
(37) from entering into transitivity alternations parallel to those of unaccusatives, as in (36).

38. a. b.

pronoun v’ v

v V pronoun v

pronoun V’ v V

R V R V

a- mo- Ø - manduʔa Ø o- mo-sê- Ø

3-CAUS-3- memory-V 3- CAUS-exit-V
‘I make him remember’ ‘He makes (him) go out’

However, both types of causativization are productive in Guaraní. It is possible that the mbo-
head is merged with the vP, rather than the VP, but the implications of this hypothesis are not
immediately evident. If a PA analysis for Guaraní verbs is to be maintained, then the structure of
the two classes of intransitive, as well as their causative alternates, should be investigated
further.

5. Conclusion and remaining questions

In this paper, I have attempted to provide an explanation of the verbal argument structure of
Guaraní that could account for the two classes of Guaraní intransitive verbs, as well as for the
resemblance between these agreement paradigms and those found in transitive verbs. Applying
the Pronominal Argument Hypothesis to Guaraní was able to explain the contrasts in intransitive
marking as reflecting the different positions of the pronominal argument relative to the verb. An
additional benefit of the PAH analysis was that the ‘free’ word order and optional status of DPs
in Guaraní could be attributed to the adjunct status of the DP in the language. Finally, a number
of other Guaraní affixes were reanalyzed as pronominal, and in doing so, a more consistent order
for Guaraní argument order emerged. The matter of the structure of causativized intransitives
remains to be explored, as well as the possibility that prolonged contact with non-PA languages
may have resulted in a linguistic shift in some dialects of Guaraní away from pronominal
arguments.

Hammink 16
References

Alexiadou, A., Anagnostopoulou, E., & Everaert, M. (2004). Introduction. In A. Alexiadou

et al. (eds) The unaccusativity puzzle : Explorations of the syntax-lexicon interface. 1-21.
Oxford; New York: Oxford University Press.

Ayala, J. V. (1993). Gramática guaraní. [Asunción]: Centro Cultural "Leopoldo Marechal"

de la Embajada de la República Argentina en el Paraguay.

Baker, M. C. (1988). Incorporation: A theory of grammatical function changing. Chicago:

University of Chicago Press.

Baker, M. C. (1996). The polysynthesis parameter. Oxford studies in comparative syntax.

New York: Oxford University Press.

Baker, Mark. (2006). On Zero Agreement and Polysynthesis. In P. Ackema, P. Brandt, M.

Schoorlemmer and F. Weerman (eds.) Arguments and Agreement. Oxford University Press,
Oxord, 289-320.

Canese, N. K. d. & Decoud Larrosa, R. (1983). Gramática de la lengua guaraní. Colección

̄
Nemitỹ . Asunción: N. Krivoshein de Canese.

Choi, J. K. (2000). [-Person] Direct Object Drop: The Genetic Cause of a Syntactic Feature
in Paraguayan Spanish. Hispania. 83 (3).

Colijn, E. (2007). Word order in Paraguayan Guaraní. Thesis (Honors)--Ohio State

University, 2007.

Folli, R., & Harley, H. (2007). Causation, obligation, and argument structure: On the nature
of little v. Linguistic Inquiry, 38(2), 197-238.

Gregores, E., & Suárez, J. A. (1967). A description of colloquial Guaraní. The Hague;
Paris: Mouton & Co.

Grimes, B. F., Grimes, J. E., & Pittman, R. S. (1996). Ethnologue Languages of the world.
Dallas, Texas: Summer Institute of Linguistics. http://www.sil.org/ethnologue/.

Grimshaw, J., & Mester, A. (1988). Light verbs and θ-marking. Linguistic Inquiry, 19(2),
205-232.

Guasch, A. (1996). El idioma guaraní: Gramática y antología de prosa y verso. Asunción:

CEPAG.

Hammink 17
Hale, K. (2003). On the significance of Eloise Jelinek's Pronominal Argument Hypothesis.
In A. Carnie, et al. (eds) Formal approaches to function in grammar: In honor of Eloise
Jelinek. Amsterdam: John Benjamins Pub.

Hale, K. L., & Keyser, S. J. (1993). On argument structure and the lexical expression of
syntactic relations. In K. Hale and S. J. Keyser (eds) The View from Building 20: Essays in
Linguistics in Honor of Sylvain Bromberger. 111-176, Cambridge, MA: MIT Press.

Hale, K. L., & Keyser, S. J. (2002). Prolegomenon to a theory of argument structure.

Cambridge, Mass.: MIT Press.

Hammink, J. E. (2006). Verb incorporation in Guaraní causative constructions. Thesis

(M.A.). University of Texas at El Paso, 2006.

Harley, H. (n.d.). A minimalist approach to argument structure. Unpublished manuscript.

Harley, H. (2002) A minimal(ish) linking theory. Handout for a talk presented at the
Maryland Minimalist Mayfest, May 16, 2002.

Jelinek, E. (1984). Empty Categories, Case, and Configurationality. Natural Language and
Linguistic Theory. 2 (1), 39-76.

Jelinek, E., & Demers, R. A. (1994). Predicates and Pronominal Arguments in Straits Salish.
Language. 70 (4), 697-736.

Jelinek, E. (2006). The pronominal argument parameter. In P. Ackema (ed.) Arguments and
agreement. 261-289. Oxford: Oxford University Press.

Jensen, Cheryl. (1987). Object-prefix incorporation in proto Tupí-Guaraní verbs. Language

Sciences, 9 (1), 45-55.

Klimov, G. A. (1974). On the character of languages of active typology. Linguistics, 131,

11-25.

Larson, R. K. (1988). On the Double Object Construction. Linguistic Inquiry. 19 (3), 335-
391.

LeSourd, P. S. (2006). Problems for the Pronominal Argument Hypothesis in Maliseet-

Passamaquoddy. Language. 82 (3).

Levin, B., & Hovav, M. R. (1995). Unaccusativity: At the syntax-lexical semantics interface.
Linguistic inquiry monographs, 26. Cambridge, Mass: MIT Press.

Mithun, M. (1991). Active/agentive case marking and its motivations. Language, 67(3)

Hammink 18
Nordhoff, S. (2004). Stative verbs and possession in Guarani. Paper presented at Syntax of
the World’s Languages Conference, Leipzig, Germany.

Perlmutter, D. (1978). Impersonal passives and the unaccusative hypothesis. Proceedings of

the Berkeley Linguistics Society, 4, 157–189.

Rose, F. (2003). Morphosyntaxe de l'émérillon. Une langue tupi-guarani de Guyane

̧
francaise . Thesis (Ph.D.): Université Lumière Lyon.

Schleicher, C. (1989). The development of the Guarani verb system. Thesis (M.A.):
University of Kansas.

Schleicher, C. O. (1998). Comparative and internal reconstruction of the Tupi-Guarani

language family. Thesis (Ph. D.): University of Wisconsin-Madison.

Trinidad Sanabria, L. (1998). Polisíntesis guaraní: Contribución para el conocimiento

tipológico de esta lengua amerindia. Asunción, Paraguay: Intercontinental Editora.

van Valin, R. D. (1990). Semantic Parameters of Split Intransitivity. Language. 66 (2), 221-
260.

Velazquez Castillo, M. (1996). The grammar of possession: Inalienability,

incorporation,and possessor ascension in Guaraní. Studies in language companion series, v.
33. Philadelphia, Pa: John Benjamins Pub.

Vieira, M. M. D. (1993). O fenômeno de não-configuracionalidade na língua asurini do

trocara. Thesis (Ph.D.): Universidade Estadual de Campinas.

Vieira, M. M. D. (1995). The Expression of Quantificational Notions in Asurini do Trocará:

Evidence Against the Universality of Determiner Quantification in E. Bach et al. (eds)
Quantification in Natural Languages. 701-720. Amsterdam: Kluwer Academic Publishers.

Vieira, M. M. D. (2001). A natureza transitiva das sentenças possessivas em Mbyá-Guaraní.

In F. Queixalós (ed.) Des noms et des verbes en tupi-guaraní : État de la question. 67-85.
München: Lincom Europa.

Hammink 19
Hammink 20
Abbreviations:

1: 1st person

2: 2nd person

3: 3rd person

ABS: absolutive case

ACC: accusative case

ART: article

CAUS: causative

DIM: diminutive

ERG : ergative case

Excl.: exclusive

FUT: future tense

Incl.: inclusive

NEG : negative

NMN : nominalizer

NOM :nominative case

PAH : Pronominal Argument Hypothesis

PL: plural

PLi: plural, inclusive

PLe: plural, exclusive

POSS: possessive

PP: postposition

PROG: progressive

Hammink 21
PTG: proto-Tupi-Guaraní

RC: reciprocal

RF: reflexive

SC: small clause

SG: singular

UO: unspecified object

Hammink 22