States, Changes of State, and ‘Feigned States’ in Paraguayan Guarani Je-/Ñe- Predicates

Abstract

Paraguayan Guarani features a pervasive pattern of aspectual polysemy whereby an underived monovalent predicate conveys both state and change of state senses, such as -o’ysã ‘be cold/get cold.’ The language also allows the derivation of monovalent predicates from causative verbs by the je-/ñe- prefix, yielding two distinct classes. The first one, called Class I, comprises predicates derived from causatives without an underived intransitive counterpart, such as -je-pe’a, from -pe’a ‘open (tr.).’ The second class, called Class II, are predicates derived from causatives with an underived intransitive counterpart, such as -je-jeko and -ñe-moro’ysã, from -jeko ‘break (tr.)’ and -moro’ysã ‘make cold,’ which are associated with the underived intransitives -jeka ‘break (intr.)’ and -o’ysã ‘be cold/get cold,’ respectively. It is shown that the aspectual polysemy pattern extends to je-/ñe-derived predicates, but that its distribution is restricted to Class I predicates. An apparent exception to this are some Class II predicates with a ‘feigned state’ interpretation, such as -ñembotavy ‘pretend to be dumb.’ The evidence reveals, however, that these are not true states, but instead activities. Furthermore, it is argued that the je-/ñe- predicates with this interpretation are not the product of a derivation process but are rather best analyzed as an independent construction.

1. Introduction

Paraguayan Guarani features a pervasive pattern of aspectual polysemy whereby a monovalent predicate conveys both state and change of state senses. A state is an aspectual class that depicts a non-dynamic (i.e., showing no change) situation. A change of state depicts the transition from a state to a different state.1 This is illustrated below, where -o’ysã means both ‘be cold’ (1a) and ‘get cold’ (1b), and -mano means ‘be dead’ (2a) and ‘die’ (2b).2

(1)	a.	Context: I take the water from the fridge. I touch it and it’s cold. I describe this.
		Ko	y	ho’ysã.
		det	water	3in.cold
		‘The water is cold.’
	b.	Context: I heat up water for mate. I forget about it and after a while it becomes cold. I describe what happened.
		Ko	y	sapy’aitépe	ho’ysã.
		det	water	right.away	3in.cold
		‘The water got cold right away.’
(2)	a.	Context: I go out to the yard and see a lifeless dog. I describe what I see.
		Pe	jagua	o-mano.
		det	dog	3ac-die
		‘That dog is dead.’
	b.	Context: I go out to the yard and I see the dog playing around. Suddenly, it shakes and drops dead. I describe what happened.
		Pe	jagua	o-mano.
		det	dog	3ac-die
		‘That dog died.’

In Califa (n.d.), this polysemy pattern is demonstrated to extend across a large number of predicates from various semantic classes (see Section 2 for details about this), thus qualifying as an instance of regular polysemy (Apresjan 1974).

In the language, it is also possible to derive monovalent predicates by attaching the je-/ñe- prefix to causative verbs (here understood as transitive verbs involving a change of state). This is exemplified below; in (3–4), lexical causatives are shown, and in (5), verbs derived by means of the causative prefix mo-/mbo-. (Ñe- is the nasal allomorph.)

		Intransitive verb		Causative verb		Je-/ñe-derived verb
(3)	a.			-pe’a ‘open’	→	-je-pe’a
	b.			-kytĩ ‘cut’	→	-ñe-kytĩ
(4)	a.	-mano ‘dead/die’		-juka ‘kill’	→	-je-juka
	b.	-jeka ‘broken/break’		-joka ‘break’	→	-je-joka
(5)	a.	-o’ysã ‘cold/get cold’	→	-mo-ro’ysã ‘make cold’	→	-ñe-mo-ro’ysã
	b.	-guapy ‘seated/sit’	→	-mbo-guapy ‘make sit’	→	-ñe-mbo-guapy

Note that three distinct classes of derived predicates can be recognized according to two different criteria to classify the causative predicate. On the one hand, a lexical–morphological criterion might be adopted that differentiates the causative verbs that lack an underived intransitive verb—as in (3)—from those that have an underived counterpart—as in (4–5). On the other hand, a purely morphological criterion might be taken that considers the formal makeup of the causative base and thus groups the verbs that are derived from an underived base—i.e., a lexical causative, as in (3–4)—and singles out those that are derived from an already derived base—i.e., a morphological causative, as in (5).

This paper shows that the aspectual polysemy pattern illustrated in (1–2) carries over to je-/ñe-derived monovalent predicates, albeit with a variable distribution across the classes in (3–5). There is agreement in the descriptive literature that the je-/ñe- predicates in (3–5) typically have reflexive or passive interpretations, that is, eventive readings (see Section 3 about this). As to the availability of stative readings, consider the following examples.

(6)	Context: I look at the house and see that the door is open. I describe that.
	Ojepe’a.
(7)	Context: I take the water from the fridge. I touch it and it’s cold. I describe this.
	#	Oñemoro’ysã.
(8)	Context: I go out to the yard and see a lifeless dog. I describe what I see.
	#	Ojejuka.

As can be seen, -jepe’a allows a stative interpretation but -ñemoro’ysã and -jejuka do not. Consistent with this, it will be shown that only the je-/ñe-derived predicates such as those in (3) are aspectually polysemous, and that the verbs such as those in (4–5) convey exclusively eventive senses. This renders the lexical–morphological criterion—i.e., the transitive base of derivation having or not an underived intransitive counterpart—the right organizing principle for the distribution of the polysemy pattern across je-/ñe-derived predicates. For expository purposes, the je-/ñe- predicates in (3) will be called Class I derived predicates and those in (4–5) Class II derived predicates.

However, an apparent exception to this distribution are some Class II predicates that have what can be termed a ‘feigned state’ interpretation, as in (9).

(9)	Aní-na	re-ñe-mbo-tavy-re-ína.
	neg-req	2sg.ac-caus-dumb-2sgac-prog
	‘Please, don’t be playing dumb.’		(Estigarribia 2020, p. 166)

Since it is formed on the basis of a morphological causative, the je-/ñe- predicate in (9) belongs to Class II. As can be seen, it expresses the sense of somebody pretending to be in the state denoted by the root predicate -tavy ‘dumb/become dumb.’ This could be taken to run counter to the claim made above that Class II predicates denote only events. Nonetheless, despite their stative flavor, it will be seen that these ‘feigned states’ are not true states but, instead, are part of another aspectual class, namely, activities. Moreover, it will be argued that the predicates such as the one in (9) are not the product of je-/ñe-derivation, but rather constitute an independent construction (Goldberg 1995, 2006) which happens to superficially coincide in form with je-/ñe-derived predicates in most cases. To my knowledge, this construction has not been examined in detail before in Paraguayan Guarani or in other Tupi-Guarani languages.

The rest of the paper is organized as follows. So as to have a clearer picture of the aspectual polysemy pattern, Section 2 provides an overview of how this is obtained with underived predicates such as those in (1–2). Section 3 sketches some of the general properties of the je-/ñe marker, especially those relevant to the predicates analyzed here. Section 4 focuses on the distribution of state and change of state senses in Class I and II predicates. Section 5 concentrates on the ‘feigned state’ interpretation, making the case for a constructional analysis. Finally, Section 6 presents the conclusions.3

2. Aspectually Polysemous Underived Predicates

As was said above, the state and change of state polysemy pattern is found with underived monovalent predicates belonging to different semantic classes. One of these classes is property concepts (PC) (Thompson 1989, 2004), which comprise the meaning types typically coded as adjectives in languages such as Spanish or English. Indeed, a large number of lexical items from the various subclasses recognized therein (Dixon 1982, 2004) prove to be aspectually polysemous in Guarani (Califa n.d.). A sample of these predicates is provided in

Table 1. Aspectually polysemous PC predicates.

Subclass	Examples
Dimension	-tuicha ‘large/enlarge (intr.),’ -puku ‘long/lengthen (intr.),’ -kyra ‘fat/become fat,’ -piru ‘thin/become thin’
Color	-pytã ‘red/redden (intr.),’ -ovy ‘blue/become blue’
Value	-ñaña ‘bad/become bad’
Age	-tuja ‘old/age’
Physical property	-aku ‘hot/become hot,’ -kã ‘dry/dry up (intr.),’ -atã ‘hard/harden (intr.)’
Human propensity	-tavy ‘dumb/become dumb,’ -ñaro ‘wild/become wild’
Speed	-pya’e ‘fast/become fast’

.

PC predicates contrast with what Beavers and Koontz-Garboden (2020) term result states (RS), within which they have proposed several subclasses. In all of these, it is possible to find abundant evidence of the polysemy pattern in Guarani as well. A sample is given in

Table 2. Aspectually polysemous RS predicates.

Subclass	Examples
Posture	-guapy ‘seated/sit,’ -ñeno ‘lying/lie down’
Physiological processes	-ke ‘asleep/fall asleep,’ -ka’u ‘drunk/get drunk’
Die verbs	-mano ‘dead/die,’ -jahoga ‘drowned/drown (intr.)’
Psychological	-pochy ‘angry/get angry,’ -vy’a ‘glad/become glad’
Entity-specific changes of state	-tuju ‘rotten/rot (intr.),’ kái ‘burnt/get burnt,’ -yku ‘melted/melt (intr.)’
Cooking	-jy ‘grilled/grill (intr.),’ -chyryry ‘fried/fry (intr.)’
Breaking	-jeka ‘broken/break (intr.),’ -soro ‘torn/tear (intr.),’ -pẽ ‘shattered/shatter (intr.)’
Bending	-karapã ‘bent/bend (intr.),’ -chaĩ ‘creased/get creased’
Miscellanea	-kirirĩ ‘quiet/shut up,’ -kañy ‘hidden/hide (intr.)’

.

Beavers and Koontz-Garboden (2020) claim that what distinguishes PC predicates—what they call ‘roots’—from RS predicates is that the latter contain a change entailment in their denotation, while the former acquire this only by combining with event structure operators. The basis for this distinction comes from two different types of semantic asymmetry.4

To begin with, while it is possible to assert a PC state and to negate its corresponding change of state—as in (10)—this cannot be done with an RS predicate without contradiction—as in (11). This is indeed predicted by the presence of the change entailment in the denotation of the latter (Beavers and Koontz-Garboden 2020, p. 64; Koontz-Garboden 2005, p. 86).

(10)	The dirt is red, but nobody reddened it.
(11)	#The glass is broken, but it never broke.	(Koontz-Garboden 2005, p. 86)

The other asymmetry stems from the scope of the modifier again, which in principle allows for two distinct readings (Dowty 1979). The first, known as the repetitive interpretation, implies the mere repetition of an event. The second, called the restitutive interpretation, involves the restoration of a state that held before the change of state occurred. Crucially, the latter is only possible when again scopes over a stative predicate. This predicts a difference which is in effect verified: while PC predicates allow both readings, RS predicates license only the repetitive one (Beavers and Koontz-Garboden 2020, p. 83). This is an expected outcome if RS predicates are assumed to be inherently eventive and only derivatively stative. (Only the restitutive readings are given in (12–13).)

(12)	Context: Kim takes a photo that is too large to use as a Facebook profile photo. She shrinks it to a more appropriate size, but thinks it does not look good. So she restores it to its original size and puts it on her personal website instead.
	Kim enlarged the photograph again.
(13)	Context: An ice-cream factory manufactures ice-cream from a package of ingredients by adding water and then freezing the result. After adding the contents of the package to water and freezing it, Kim lets it melt into a liquid state.
	#Kim melted the ice-cream again.	(Beavers and Koontz-Garboden 2020, p. 85)

Note that in (12), the restitutive reading is obtained with PC change of state enlarge, while in (13), it does not with RS change of state melt.

These two asymmetries are indeed attested in Guarani. Consider the examples in (14).

(14)	a.	Ko’ãga		tape	i-puku	ha	siempre	upéicha	va’ekue.
		now		road	3in-long	coord	always	like.that	pas
		‘Now the road is long and it’s always been like that.’
	b.	#	Ko’ãga	ovetã	o-jeka	ha	siempre	upéicha
			now	window	3ac-break	coord	always	like.that
			o-ime	va’ekue.
			3ac-be	pas
			‘Now the window is broken and it’s always been like that.’

It can be seen above that the state of a PC predicate such as -puku ‘long/lengthen’ can be asserted (14a) while negating its change of state, but that this is not possible with an RS predicate such as -jeka ‘broken/break’ (14b).5

Now consider the data in (15–16). The marker -jey is the equivalent of again (it is glossed ‘repetitive’ here, following Gynan 2017)

(15)	a.	Repetitive context: We buy bread and leave it on the table. It hardens. We warm it in the oven and it becomes soft. However, we leave it on the table and forget it’s there so it hardens for a second time.
		Pan		hatã-jey.
		bread		3in.hard-rep
		‘The bread hardened again.’
	b.	Restitutive context: We grab a steel plaque. Unwittingly, we leave it by the fire and it softens. We remove it from near the heat and a while later it’s hard again.
		Acero		hatã-jey.
		steel		3in.hard-rep
		‘The steel is hard again.’
(16)	a.	Repetitive context: I take out an ice tray from the freezer. I leave it on the counter and it soon melts. I put it back into the freezer and it freezes. I take it out again, leave it on the counter, and it melts for a second time.
		Hielo		hyku-pa-jey.6
		ice		3in.melt-comp-rep
		‘The ice melted again.’
	b.	Restitutive context: We pour water in the ice tray, put it into the freezer, and it freezes. We take it out and it melts.
		#	Hielo	hyku-pa-jey.
			ice	3in.melt-comp-rep
			‘The ice is melted again.’

This yields the same contrast regarding restitutive readings shown for English in (12–13). While a PC predicate such as -atã ‘hard/harden’ allows both readings, that is, a repeated hardening event (15a) and a restored state of being hard (15b), an RS predicate such as -yku ‘melt’ gives rise only to the repetitive reading of water becoming frozen for a second time (16a). The restitutive reading where the water goes back to the state prior to becoming frozen—i.e., being liquid—is barred for this predicate under -jey modification (16b).

In sum, the state and change of state polysemy pattern extends to both classes, which are identified on independent grounds. As for the je-/ñe-derived predicates, it will be seen below that those that prove to be polysemous pattern with RS predicates.

Finally, a word must be said about a key property of the morphology of Guarani verbs (and of other Tupi-Guarani languages as well). As might have been gleaned from the examples presented thus far, the language shows a split for argument marking in intransitive verbs (the active vs. inactive paradigms). There is wide consensus in the literature that the split is semantically based, though without agreement as to the exact parameters that condition it. For example, Mithun (1991) argues that the divide is explained in terms of lexical aspect, with inactively marked verbs denoting states and actively marked ones denoting events. Nevertheless, as the pervasiveness of the aspectual polysemy pattern shows, this is untenable (see Tonhauser 2006, p. 257 for other arguments against this). Velázquez-Castillo (1996) and Estigarribia (2020), for their part, propose participant control and volition as relevant properties.

The data presented here show that all PC predicates are inactively marked. RS predicates, in contrast, are not uniform. While the majority are actively marked, a considerable number of arguably RS predicates are inactively marked: -yku ‘melted/melt’ (exemplified in (16)), -asy ‘sick/get sick,’ -kane’õ ‘tired/get tired,’ -pochy ‘angry/get angry,’ -kaigue ‘bored/get bored,’ -kuerái ‘fed-up/get fed-up,’ -atĩ ‘white-haired/become white-haired,’ -tuju ‘rotten/become rotten,’ -ykue ‘wet/become wet,’-chaĩ ‘creased/become creased.’ All je-/ñe- predicates, for their part, are invariably actively marked and, as will be shown, pattern with RS predicates. These facts suggest that the characterization of the PC vs. RS classes could contribute to the understanding of the semantic basis of the split, especially for inactive predicates. However, I pursue this question no further here (see Califa n.d., for more on this).

3. The Je-/Ñe- Marker

Before moving on to the analysis of the derived predicates, it is instructive to provide a concise and general characterization of the je-/ñe- marker. One of the main functions of this prefix is to derive monovalent predicates from bivalent ones. In keeping with this, it has been variously analyzed as a reflexive (Gregores and Suárez 1967), a middle (Velázquez-Castillo 2008; Estigarribia 2017a), or an agent-demoting marker (Estigarribia 2020). The latter characterization has been adopted here.7

As was said in Section 1, je-/ñe- typically gives rise to two distinct diathetic readings (Zúñiga and Kittilä 2019), a reflexive and a passive-like interpretation. Indeed, the same form can be ambiguous between the two readings, as illustrated in (17) with the lexical causative -kytĩ ‘cut.’ It should be kept in mind that, regarding the passive reading, the agent obligatorily remains implicit and can in no way appear overtly (Tonhauser 2017, p. 208; Velázquez-Castillo 2008, p. 390).

(17)	O-ñe-kytĩ.
	3ac-agd-cut
	‘(S)he was cut.’/‘(S)he cut herself/himself.’	(Estigarribia 2020, p. 210)

Some Class I derived causatives can likewise be interpreted as a spontaneous, non-agentive occurrence, very much like the anticausatives of languages such as Spanish (Haspelmath and Müller-Bardey 2004).

(18)	Okẽ	o-je-pe’a.
	door	3ac-agd-open
	‘The door opened.’ (Also: ‘the door was opened’ or ‘someone opened the door.’)

The same diathetic ambiguity is also obtained with bivalent non-causative verbs, as shown in (19) with -japi ‘shoot.’

(19)	Tomás	o-je-japi.
	Tomás	3ac-agd-shoot
	‘Tomás was shot.’/‘Tomás shot himself.’		(Velázquez-Castillo 2008, p. 389)

The prefix also combines with monovalent verbs, the effect of which is to convey an impersonal or generic reading.

(20)	O-je-jeroky.
	3ac-agd-dance
	‘There is dancing.’	(Estigarribia 2020, p. 208)

As Estigarribia’s (2020) characterization underscores, what all these uses have in common is that the agent is somehow demoted, by leaving it implicit as an unspecified or indefinite participant—as in the passive or generic/impersonal interpretations—by suppressing it altogether—as in the case of the anticausative reading—or by making it coincide with the patient—as in the reflexive reading. From the aspectual point of view, these all constitute eventive readings.8

4. States and Changes of State in Derived Predicates

4.1. Class I Derived Predicates

Since Class I predicates are formed on the basis of lexical causatives, a few words about these are in order. In Guarani, lexical causatives come in much smaller numbers than morphological causatives, which are by far more common in the lexicon (Velázquez-Castillo 2002). Velázquez-Castillo (2002, p. 512) distinguishes three types of lexical causatives according to the formal relation they establish with their intransitive counterparts. The first two types have underived intransitive counterparts and, thus, fall into Class II, to be reviewed in Section 4.2. The relevant type here comprises the causatives that lack an underived intransitive counterpart. A non-exhaustive list is provided in

Table 3. Lexical causatives with no underived intransitive counterpart.

Verb	Meaning
-pe’a	‘open’
-mboty	‘close’
-kytĩ	‘cut’
-haimbe’e	‘sharpen’
-monde	‘dress’
-jokua	‘tie’
-hundi	‘lose/break (e.g., a gadget)’
-pyso	‘unfold’
-poka	‘wring’
-karãi	‘scratch’
-mbyai	‘damage, ruin’
-juvy	‘strangle’
-ñotỹ	‘bury’
-jahói	‘cover’
-johéi	‘wash’
-mbichy	‘toast’
-mondýi	‘scare’
-picha	‘offend, annoy’
-ko’õi	‘offend’

. (The mo-/mb- formant that can be discerned in some cases is lexicalized.).

In order to obtain their intransitive counterparts, these verbs must necessarily resort to je-/ñe derivation. The derived predicates resulting from this process are the members of what are here called Class I predicates.

As was explained in Section 3, two of the interpretations available for the je-/ñe- derived predicates from these lexical causatives are the reflexive and passive readings, as attested in (17), repeated below as (21).

(21)	O-ñe-kytĩ.
	3ac-agd-cut
	‘She/he was cut.’/’She/he cut her/himself.’	(Estigarribia 2020, p. 210)

Note that these are by definition eventive readings. The crucial piece of evidence to know if they are aspectually polysemous is whether they also convey a stative sense. Although some isolated remarks to this effect about specific predicates can be found in the literature (e.g., Dietrich 2017, p. 175 on jepe’a), no systematic data about its extent are available. The research conducted on a wide array of lexical items for this paper reveals that Class I predicates can have a stative interpretation too. A number of representative examples are given in (22–26).

(22)	Context: Horacio bought a shirt to wear at a party. However, when he gets home, he puts it on and sees that the shirt has a cut on one it of sleeves.
	Pe	kamisa	o-ñe-kytĩ.
	det	shirt	3ac-agd-cut
	‘The shirt has a cut on it.’9
(23)	Context: I’m going to a party with Verónica tonight. She had the foresight to get ready in the afternoon and is now fully dressed. I describe what I see.
	Verónica		o-ñe-monde.
	Verónica		3ac-agd-dress
	‘Verónica is dressed.’
(24)	Context: I want to listen to the radio but it won’t turn on. I realize it’s not working. I describe that.
	Radio		o-ñe-mbyai.
	radio		3ac-agd-damage
	‘The radio is broken.’
(25)	Context: Valeria and I argued. She said very hurtful and offensive things to me. I describe how I feel.
	A-je-picha.
	1sg.ac-agd-offend
	‘I’m offended.’
(26)	Iñ-elegante-eterei			ha	o-ñe-mondýi.
	3in-elegant-sup			coord	3ac-agd-scare
	‘She’s very elegant and she’s scared.’					(Estigarribia 2017b, p. 332)

The derived predicates of the rest of the verbs in Table 3 have likewise been documented to have stative senses apart from their eventive senses.

Table 4. Aspectually polysemous Class I derived predicates.

Predicate	Stative Reading	Change of State Reading
ojepe’a	‘it’s open’	‘it opened’
oñemboty	‘it’s closed’	‘it closed’
oñehaimbe’e	‘it’s sharpened’	‘someone sharpened it’
ojejokua	‘it’s tied’	‘someone tied it’
oñehundi	‘it’s lost/broken’	‘it became lost/broken’
ojepyso	‘it’s unfolded’	‘someone unfolded it’
ojepoka	‘it’s wrung’	‘someone wrung it’
oñekarãi	‘it’s scratched’	‘someone scratched it’
ojejuvy	‘she/he’s strangled’	‘someone strangled her/him’
oñeñotỹ	‘it’s buried’	‘someone buried it’
ojejahói	‘she/he’s covered’	‘she/he covered her/himself’
ojejohéi	‘it’s washed’	‘someone washed it’
oñembichy	‘it’s toasted’	‘it got toasted’
oñeko’õi	‘she/he’s offended’	‘she/he took offence’

lists the possible stative and change of state readings for these predicates in third-person form. (For the change of state interpretations, only one illustrative diathesis is given—spontaneous occurrence, reflexive, passive, or some other.)

It follows, then, that Class I je-/ñe-derived predicates are aspectually polysemous as they convey both state and change of state senses, thus patterning with the underived predicates seen in Section 2. It was shown, however, that there are two clearly distinguished semantic classes among the latter, namely, PC predicates and RS predicates. Which of these do Class I predicates behave like?

Recall that according to Beavers and Koontz-Garboden (2020), the difference between these two classes resides in that RS predicates come with an entailment of change in their denotation. This is what accounts for the asymmetries evidenced by the tests in (14–16). Regarding the first test, it was seen that RS predicates are not felicitous with a continuation that negates the change of state. Consider the examples in (27–29) with -ñemonde ‘be dressed/get dressed,’ -jepe’a ‘be closed/close (intr.),’ and -ñembyai ‘be damaged/get damaged.’

(27)	#	Ko’ãga	Brenda	o-ñe-monde	ha	siempre	upéicha
		now	Brenda	3ac-agd-dress	coord	always	like.that
		o-ime	va’ekue.
		3ac-be	pas
		‘Now Brenda is dressed and she’s always been like that.’
(28)	#	Ko’ãga	okẽ	o-je-pe’a	ha	siempre	upéicha
		now	door	3ac-agd-open	coord	always	like.that
		o-ime	va’ekue.
		3ac-be	pas
		‘Now the door is open and it’s always been like that.’
(29)	#	Ko’ãga	radio	o-ñe-mbyai	ha	siempre	upéicha
		now	radio	3ac-agd-damage	coord	always	like.that
		o-ime	va’ekue.
		3ac-be	pas
		‘Now the radio is broken and it’s always been like that.’

It can be here seen that, like RS predicates, the states of these derived verbs cannot be combined with a continuation negating the corresponding change of state.

The other test involved the availability of repetitive and restitutive readings under again modification, with RS predicates licensing the former but disallowing the latter. Look at the example in (30) with -ñekytĩ ‘be cut/get cut.’

(30)	a.	Repetitive context: We put up a clothesline in our building’s terrace. One day, we went up and saw somebody had cut it off. We tied it together and put it back up. However, the next day we saw somebody had cut it a second time.
		Sã		o-ñe-kytĩ-jey.
		clothesline		3ac-agd-cut-rep
		‘Somebody cut the clothesline again.’
	b.	Restitutive context: We put up a clothesline in our building’s terrace. One day, we went up and see somebody had cut it off and left it in two pieces on the floor. We tied them together and put the fixed clothesline back up. However, as we tied the two pieces too loosely, the knot that was keeping them together came untied and they fell onto the floor.
		#	Sã	o-ñe-kytĩ-jey.
			clothesline	3ac-agd-cut-rep

Notice that, when modified by -jey ‘again,’ the predicate is fine for the repetitive reading of a second cutting event (30a), but it is not felicitous in a context where the result state of the cutting event is restored (30b)—i.e., the clothesline being in two pieces again. Here, -ñekytĩ aligns with RS predicates.

An important qualification needs to be made here concerning this parallelism between Class I predicates and RS predicates. Recall that, as Beavers and Koontz-Garboden’s (2020) make clear, the asymmetries revealed by these tests are indicative of two types of roots—in the terminology of this paper, underived predicates. The first type refers to roots that have an element of change introduced in their event structures only via operators, while the second type are roots that inherently possess an entailment of change in their denotation. It must be stressed that their notion of root does not refer to a morphological unit but rather to the lexical-semantic constant of an event structure (Beavers and Koontz-Garboden 2020, p. 9). Class I derived predicates, by contrast, do not seem to be roots in this sense. This is because they are not only morphologically derived but, crucially, also event-structurally derived. Their derivation takes a causative verb as its input and yields a non-causative, aspectually polysemous verb. This means that the sort of behavior illustrated in (27–30) is not, in principle, informative about the nature of the roots in their event structures as it would be for uncontroversially RS roots such as the ones presented in Section 2. The question about the type of root involved in Class I predicates is left open here.

In sum, it is concluded that the state and change of state polysemy pattern extends to Class I derived predicates, that is, the je-/ñe- predicates derived from causatives without intransitive counterparts. The next section analyzes Class II derived predicates.

4.2. Class II Derived Predicates

Class II predicates are derived from lexical or morphological causative verbs that have an underived intransitive verb. The lexical causatives fall into two small-sized groups in terms of the formal relation they establish with their intransitive counterparts (Velázquez-Castillo 2002). The first are a few causative and intransitive pairs that feature vowel alternation; the second group comprises a few suppletive pairs; both are shown in

Table 5. Lexical causatives with underived intransitive counterparts.

	Intransitive	Causative
Vowel alternation pairs	-jeka ‘broken/break’	-joka ‘break’
	-jera ‘untied/come untied’	-jora ‘untie’
	-jeko ‘leaning/lean against X’	-joko ‘stop X physically’
Suppletive pairs	-mano ‘dead/die’	-juka ‘kill’
	-kái ‘burnt/get burnt’	-hapy ‘burn’
	-sẽ ‘get out’	-nohe ‘to take out’

. Note that the intransitives of these two groups are members of the class of underived predicates described in Section 2 and are accordingly aspectually polysemous (in fact, -jeka was illustrated in (14b)). An exception to this is -sẽ, which has solely the eventive sense (that is, it means ‘get out’ but not ‘be outside’).

Morphological causatives, by contrast, form a much larger class. Guarani possesses three causative morphemes: mbo-/mo-, guero-/ro-, and -uka (Estigarribia 2017a, 2020; Velázquez-Castillo 2002). The most productive is the prefix mbo-/mo-, which accounts for the majority of the causative verbs in the lexicon. Indeed, most of the underived predicates presented in Section 2 can be causativized by means of this prefix. A few examples are given in (31): a–b from PC predicates, c–d from RS predicates. (Mo- is the nasal allomorph.)

(31)	a.	-kã ‘dry/dry up (intr.)’	→	-mo-kã ‘dry (tr.)’
	b.	-puku ‘long/lengthen (intr.)’	→	-mbo-puku ‘lengthen (tr.)’
	c.	-ke ‘asleep/fall asleep’	→	-mo-nge ‘make someone sleep’
	d.	-chyrry ‘fried/become fried’	→	-mbo-chyryry ‘fry’

As said above, the monovalent predicates derived by means of the je-/ñe- prefix from the causatives in Table 5 and the ones exemplified in (31) are what are termed Class II predicates here, that is, predicates derived from a causative with an underived intransitive counterpart. Note that in the case of the morphological causatives, this translates into the same root—in a purely morphological sense—figuring in two types of monovalent predicates, an underived and a derived one (e.g., -kã and -ñemokã, -ke and -ñemonge).

Like other je-/ñe- predicates, Class II predicates can be interpreted as reflexives or passives, as exemplified in (32–33) with -ñemopotĩ ‘clean, wash oneself,’ -ñemboguapy ‘be made to sit down’ and -jejoko ‘be held in place.’

(32)	E-gueru	chéve	y-mi	ikatu-haguã-icha
	imp-bring	1sg.dat	water-dim	be.able-for-like
	a-ñe-mo-potĩ.
	1sg.ac-agd-caus-clean
	‘Bring me a little water so I can wash myself.’
	(Lustig 2005, p. 76; cited in Estigarribia 2017a, p. 78)
(33)	Juan	o-ñe-mbo-guapy	ha		o-je-joko.
	Juan	3ac-agd-caus-sit	coord		3ac-agd-hold.in.place
	‘Juan was made to sit down and was held in place.’
	(Correa 1981, p. 14; cited in Velázquez-Castillo 2002, p. 515)

The key question, of course, is whether these predicates can also be interpreted statively.

It must first be said that the je-/ñe- predicates that are derived from the lexical causatives of the type in Table 5 and the morphological causatives with an RS base predicate do not have a stative interpretation. Consider the examples in (34–36).

(34)	Context: I walk into my bedroom and I see the window pane is missing a piece.
		O-jeka.
		3ac-break
		‘It’s broken.’
	?	O-je-joka.
(35)	Context: I go out to the yard and see a lifeless dog. I describe what I see.
		Pe	jagua	o-mano.
		det	dog	3ac-die
		‘That dog is dead.’
	#	Ojejuka. (Speaker’s comment: “OK if it’s evident someone killed it.”)
(36)	Context: I walk into the room and I see a child sitting in a chair. I describe what I see.
		O-guapy.
		3ac-sit
		‘He’s sitting.’
	#	O-ñe-mbo-guapy.

In (34), the underived RS verb ojeka ‘it’s broken’ is felicitous, while the acceptability of the predicate ojejoka ‘it was broken’—derived from causative joka ‘break’—is doubtful at best. In (35), the contrast is crisper: while underived RS omano ‘it’s dead’ is perfectly fine, derived ojejuka ‘it was killed’ is unambiguously infelicitous for the stative context. In (36), the contrast between the two forms is also very clear: underived RS oguapy ‘he’s sitting’ is fine but derived oñemboguapy ‘he was made to sit’ is not.

With the derived verbs from PC predicates, the story could arguably be different. Here, it is necessary to bring up a point about the distinction between PC and RS predicates not mentioned before. While RS predicates by definition yield only result state readings, PC predicates are not restricted in this regard. That is, they can denote either an inherent state that holds with no prior event giving rise to it, or a state that results from a prior event. In some languages, this has a morphological correlate, with the result state form being somehow more complex than the one expressing the inherent state. This is exemplified by the English adjectives long and lengthened in (37).

(37)	a.	Sandy’s shirt has long sleeves.
	b.	Sandy’s shirt has lengthened sleeves.	(Koontz-Garboden 2005, p. 87)

As Koontz-Garboden (2005, p. 87) explains, while in (37a) Sandy’s shirt’s sleeves are described to be inherently long, in (37b) they are long as a result of some lengthening process.

Recall that, with PC predicates, the same root—again, in a strictly morphological sense—could appear in two different monovalent predicates, an underived and a derived one, e.g., -kã and -ñemokã. The distinction between the two types of states pertaining to PC predicates described above—inherent states and result states—and the corresponding morphological asymmetry illustrated with English adjectives in (37) suggests that a division of labor of sorts could be postulated for the two forms of the same root in Guarani. In other words, is it possible that the underived predicate expresses the inherent state and the je-/ñe-derived predicate the result state of a given PC meaning? In the case of ‘dry,’ this analysis would predict that -kã is ‘inherently dry’ and -ñemokã is ‘dry as a result of having (been) dried.’

However, this is not what the evidence shows. Consider the following examples.

(38)	a.	Context: We go on holiday to the desert. We see it’s a very dry place. I describe that.
			I-kã-iterei.
			3in-dry-sup
			‘It’s very dry.’
		#	O-ñe-mo-kã.
	b.	Context: A child jumped into the pool and when he got out his mother dried him off with a towel. Now he’s completely dry. I describe that.
			Mitã	i-kã-mba	porque	i-sy	o-mo-kã
			child	3in-dry-comp	because	3pos-mother	3ac-caus-dry
			chupe.
			him
			‘The child is completely dry because him mother dried him off.’
		#	Mitãoñemokã porque isy omokã chupe.

(39)	a.	Context: The road to school is very long. I describe that.
			Tape	i-puku-iterei.
			road	3in-long-sup
			‘The road is very long.’
		#	Tape oñembopukuiterei.
	b.	Context: The paved stretch of the road was very short, but the city’s workmen lengthened it. Now it’s very long. I describe that.
			Tape	i-puku-iterei	porque	mba’apoha-kuéra	o-mbo-puku.
			road	3in-long-sup	because	worker-pl	3ac-caus-long
			‘The road is very long because the workmen lengthened it.’
		#	Tape oñembopuku porque mba’apoharakuéra ombopuku.

(40)	a.	Context: The soil in Misiones [Argentinian province] is red. I describe that.
			Misiones	yvy		i-pytã.
			Misiones	soil		3in-red
			‘The soil of Misiones is red.’
		#	Misiones yvy oñemopytã.
	b.	Context: The kitchen wall was white, but Carlos decided to paint it red. Now it’s red. I describe that.
			Pare		i-pytã	porque	Carlos	o-mo-pytã.
			wall		3in-red	because	Carlos	3ac-caus-red
			‘The wall is red because Carlos painted it red.’
		#	Pare oñemopytã porque Carlos omopytã.

The contexts in (38a), (39a), and (40a) describe inherent states that do not presuppose a previous event, while the contexts in (38b), (39b), and (40b) depict states resulting from events. In both cases, the form that expresses the two types of state is the underived predicate. Crucially, the derived predicate is infelicitous for the resultative state readings, so no division of labor between the two forms can be postulated in terms of stative interpretations.

Finally, in Section 1, it was mentioned that some Class II predicates have an alternative ‘feigned state’ interpretation. Example (9) is repeated below as (41).

(41)	Aní-na	re-ñe-mbo-tavy-re-ína.
	neg-req	2sg.ac-caus-dumb-2sg.ac-prog
	‘Please, don’t be playing dumb.’		(Estigarribia 2020, p. 166)

Far from being idiosyncratic, this is a very common interpretation of some Class II predicates. Nearly any predicate that can conceivably be associated with a specific attitude, demeanor, or pattern of behavior can give rise to a ‘feigned state’ interpretation. A non-exhaustive sample is given in

Table 6. ‘Feigned state’ predicates.

Verb	Meaning
-ñemoatã	‘pretend to be tough’ (lit. ‘hard’)
-ñemokangy	‘pretend to be weak’
-ñembotuicha	‘pretend to be a big-shot’ (lit. ‘big’)
-ñembopochy	‘pretend to be angry’
-ñemonge	‘pretend to be asleep’
-ñemokañy	‘pretend to be lost’
-ñembopiru	‘pretend to be thin’
-ñemongyra	‘pretend to be fat’ (e.g., by wearing a costume)
-ñembohasy	‘pretend to be sick’
-ñemoñaña	‘pretend to be mean’
-ñemokane’õ	‘pretend to be tired’
-ñemonga’u	‘pretend to be drunk’
-ñembovy’a	‘pretend to be happy’
-ñembopya’e	‘pretend to be fast’
-ñembotuja	‘pretend to be old’
-ñembovai	‘pretend to be ugly’
-ñemoporã	‘pretend to be pretty’

. Moreover, in some cases, this interpretation seems to be the most prominent one. For example, -ñembotavy also has the passive interpretation ‘be deceived’ (literally, ‘be made a fool of’). However, when presented in isolation, speakers very often report that the first interpretation that comes to mind is ‘pretend to be dumb.’ Furthermore, in dictionaries, the ‘feigned state’ reading tends to be listed along with the reflexive or passive one, not unfrequently as the first sense.

As can be expected, this sense is restricted to animate subjects. Indeed, the je-/ñe- predicates with inanimate subjects give rise only to passive readings, as in (42), or to spontaneous occurrences, as in (18).

(42)	a.	Cemento	o-ñe-mo-atã.
		concrete	3ac-agd-caus-hard
		‘The concrete was hardened.’
	b.	Koty	o-ñe-mbo-tuicha.
		bedroom	3ac-agd-caus-large
		‘The bedroom was enlarged.’
	c.	Che-róga	o-ñe-mo-porã.
		1sg.pos-house	3ac-agd-caus-pretty
		‘My house was embellished.’

What is of crucial importance here is whether ‘feigned states’ are genuine states. If this proves to be the case, then the subset of Class II predicates that have this interpretation could be said to be aspectually polysemous like Class I predicates. To examine this hypothesis, it is necessary to apply stativity tests.

To start with, states have been claimed to be incompatible with the progressive aspect (Comrie 1976). In Guarani, the progressive is expressed by means of the adverb hína (Estigarribia 2017a, 2020; Gregores and Suárez 1967).

(43)	Yvy	hykue	hína	kuri.
	earth	3in.wet	prog	pas
	‘The earth was getting wet.’				(Gregores and Suárez 1967, p. 155)

Nevertheless, Tonhauser (2006, p. 273) argues that the Guarani progressive is not restricted to dynamic predicates, as is shown in the following example.

(44)	O-ĩ	hína	vakuna	local-pe.
	3ac-be	prog	vaccinations	store-in
	‘There are vaccinations in the store now.’				(Tonhauser 2006, p. 274)

Tonhauser further observes that hína is not only compatible with the stative predicate oĩ ‘there is/are’ but also adds a sense of “’immediate relevance’ or ‘temporaryness’ [sic.]” (Tonhauser 2006, p. 274).10 With dynamic predicates, hína gives rise to an interpretation of ongoingness, in line with the English progressive. This characterization of hína predicts that if ‘feigned states’ are states, a sense of ‘immediate relevance’ or ‘temporariness’ should be obtained. If, on the other hand, they are dynamic, an ongoing reading should arise.

First, consider the following examples with je-/ñe- predicates conveying a ‘feigned state’ interpretation and hína.

(45)	a.	Nicolás	o-ñe-mo-nga’u	hína.
		Nicolás	3ac-agd-caus-drunk	prog
		‘Nicolás is pretending to be drunk.’
	b.	María	o-ñe-mbo-tavy	hína.
		María	3ac-agd-caus-dumb	prog
		‘María is pretending to be dumb.’

Notice that the examples seem to bring forth a distinct sense of ongoingness. However, the soundness of this evidence hinges on the existence of a principled distinction between ongoingness and ‘temporariness.’ By definition, a temporary eventuality should not be compatible with an expression of temporal permanence; this should not be the case with an ongoing eventuality. Two expressions of permanence in Guaraní are siempre ‘always’ (the Spanish borrowing is used) and manterei ‘constantly, at all times.’ This predicts that (44) should be odd with them but (45) should be fine. Consider (46–47).

(46)	Siempre/Manterei	o-ĩ	hína	vakuna	local-pe.
	always/constantly	3ac-be	prog	vaccinations	store-in
	‘There are indeed vaccinations at the store at all times/constantly.’

(47)	a.	Siempre/Manterei	Nicolás	o-ñe-mo-nga’u	hína.
		always/constantly	Nicolás	3ac-agd-caus-drunk	prog
		‘Nicolás is always/constantly pretending to be drunk.’
	b.	Siempre/Manterei	María	o-ñe-mbo-tavy	hína.
		always/constantly	María	3ac-agd-caus-dumb	prog
		‘María is always/constantly pretending to be dumb.’

The interpretations of ongoingness for the ‘feigned state’ predicates under the modification of siempre and manterei are acceptable; what is understood in (47) is that there is a habit of Nicolás and María pretending to be drunk and dumb, respectively. The sequence oĩ hína in (46) is also acceptable, but with a significant difference. As has been observed in the descriptive literature, hína is not always a progressive marker, but sometimes has an emphatic function (Estigarribia 2020, p. 225). As a matter of fact, this is the interpretation that speakers report for oĩ hína under the modification of siempre and manterei in (47), which they translate as Spanish “hay nomás” (rendered as English ‘there are indeed’). Importantly, they note the ‘temporary state’ reading in (44) is not available here, as is expected when modified by an expression of permanence. Therefore, while oĩ ‘there is’ displays the properties of a state, je-/ñe- predicates with a ‘feigned state’ reading as in (45) do not.

Another stativity diagnostic that has been proposed is the incompatibility of states with imperatives (Binnick 1991; Lakoff 1966), as shown in English with *Know the answer! (Binnick 1991, p. 174).11 In Guarani, the imperative with active verbs—as is the case of je-/ñe- predicates—in the second singular person is signaled by the prefix e-, as in e-karu imp- eat ‘eat’ (Estigarribia 2020, p. 170; Gregores and Suárez 1967). If the ‘feigned state’ is truly stative, then the imperative should be at least awkward in a context that describes such a situation. Yet, this is not what the facts reveal.

(48)	Context: Two little brothers are playing in their bedroom late at night when they are supposed to be fast asleep. Suddenly, one of them hears their mother walking down the corridor. One of the kids tells his brother to get in bed and close his eyes so their mother thinks they’re asleep.
	E-ñe-mo-nge!
	imp-agd-caus-sleep
	‘Pretend to sleep!’

The imperative here is perfectly fine with the ‘feigned state’ interpretation, which suggests that these are not genuine states (also consider the example in (41) with the negative imperative).12

Finally, some adverbials have been claimed to be incompatible with states. A case in point are bounded adverbials such as in two minutes, which are used only with telic events and are, thus, not possible with states (Dowty 1979; Vendler 1957). As a matter of fact, Moens and Steedman (1988, p. 21) single them out as one of the morphosyntactic contexts that coerce state predicates into a telic interpretation. In Califa (2018), I have argued that these adverbials indeed coerce the change of state reading of polysemous predicates of the underived class presented in Section 2, as is illustrated in (49–50) ((50) repeats (1b) without its context).

(49)	Ko	tomate	i-pytã	ára	kõi-me.
	det	tomato	3in-red	day	two-in
	‘The tomato became red in two days.’
(50)	Ko	y	sapy’aitépe		hoy’sã.
	det	water	right.away		3in.cold
	‘The water got cold right away.’					(Califa 2018, p. 87)

In (49–50), polysemous -pytã ‘red/become red’ and -oy’sã ‘cold/get cold’ can be interpreted only as events, not as states. Therefore, bounded adverbials such as ára kõime ‘in two days’ and sapy’aitépe ‘right away, in a bit’ display the same restriction against states identified in other languages.

Turning back to ‘feigned states,’ what is predicted is that if these are stative, they should not acceptably combine with bounded adverbials and, conceivably, would be coerced into a different reading. However, again, this prediction is not borne out.

(51)	Daniel	sapy’aitépe	o-ñe-mo-nge.
	Daniel	right.away	3ac-agd-caus-sleep
	‘Daniel pretended to be asleep right away.’

The ‘feigned state’ sense in (51) is indeed preserved under the modification of bounded adverbial sapy’aitépe ‘right away.’

On balance, all the data just analyzed point unambiguously to ‘feigned states’ not being true states. (Ironically, this renders the ‘feigned state’ term a bit of a misnomer from the aspectual point of view, though it is still useful as a shorthand since it is descriptive of its semantic gist.) What is the alternative, then? Are these changes of state, just like the other interpretations of Class II predicates? The general flavor of ‘feigned states’ that the examples afford does not agree with this, however. In all cases, the overarching meaning that can be gleaned is that someone is purposefully putting up an appearance or behavior to trick others. No sense of change seems to be part of what these ‘feigned states’ convey.

A decisive piece of evidence that helps clarify the question is given below.

(52)	Pyhare	guive	María	o-ñe-mbo-pochy.
	night	since	María	3ac-agd-caus-angry
	‘María has been pretending to be angry since last night.’

Here, the durative adverbial pyhare guive ‘since last night’ yields an interpretation of María having started to pretend to be angry last night and keeping it going uninterrupted until utterance time. The sentence is certainly not open to an interpretation in which María has been getting angry time and time again since last night, which would be expected of a change of state.

The aggregate of evidence reviewed in the preceding paragraphs shows that ‘feigned states’ are dynamic and durative. This constellation of properties corresponds to an aspectual class not mentioned before, that is, activities (Dowty 1979; Smith 1997; Van Valin 2005; Van Valin and LaPolla 1997; Vendler 1957). At this point, it might be asked whether this means that the Class II predicates with a ‘feigned state’ interpretation could also be said to be aspectually polysemous but, unlike Class I predicates, involving a different pair of aspectual classes as part of their interpretation, namely, changes of state and activities. However, as will be shown in the next section, ‘feigned states’ are not a reading stemming from Class II predicates as they have been defined here. In fact, it will be seen that the reading does not arise derivationally but is rather the semantic component of an independent construction.

To sum up this section, it can be concluded that, unlike Class I predicates, Class II predicates are not aspectually polysemous, but only give rise to change of state readings, either reflexive or passive. Thus, the principle that explains the distribution of the polysemy pattern across derived predicates is that this extends only to je-/ñe-derived verbs whose transitive base lacks an underived intransitive. This is understandable insofar as Class I je-/ñe- predicates are the only monovalent forms corresponding to their causatives.

5. The ‘Feigned State’ Construction

In this section, it will be argued that the best analysis for the je-/ñe- predicates that convey a ‘feigned state’ interpretation is that they constitute a construction in its own right, independent of the je-/ñe-derived predicates reviewed above. When a wider range of evidence is considered, the constructional account turns out to make the right predictions while the derivational account fails to capture the facts correctly.

A construction is a pairing of meaning and form (Goldberg 1995, 2006). A key question is what should be taken into account to recognize an independent construction as such. Although the different models vary in how they approach this, Goldberg’s (1995) influential criterion is particularly clear.

(53)	C is a construction iffdef C is a form–meaning pair <F1, S1> such that some aspect of F, or some aspect of S, is not strictly predictable from C’s component parts or from other previously established constructions.
	(Goldberg 1995, p. 4)

Thus, total or partial lack of compositionality is the hallmark of a construction. Note this is exactly the opposite of what the output of a derivation is supposed to be, where both the formal and semantic features of the linguistic unit should be predictable from its component parts.

These characterizations are undoubtedly gross simplifications. Rather, they are meant to serve as background assumptions to assess whether the ‘feigned state’ sense can be attributed to the je-/ñe-derived predicates presented above or should be seen as the semantic component of an independent construction. Below, two representations are provided of each analysis for this sense. The constructional schema in (54) is loosely based on the ones found in Goldberg (1995). In (55), a representation of the derivation is given, following the simple diagrams for the derivations in (3–5).

(54)	‘Feigned State’ Constructional Schema
	Meaning	Pretend to ‘predicate’
	Form	-ÑEM(B)O + Verbintransitive

(55)	Je-/Ñe- Derivation
	(Verbintransitive	→)	Verbcausative	→	je-/ñe- Verb

The constructional schema includes the string -ñem(b)o and an open slot for the insertion of the verb, which carries the specification “intransitive.” The derivation, by contrast, involves a rule that takes a causative verb as its input to yield a je-/ñe- verb. There is an optional previous rule—shown in brackets—that applies to underived verbs that derive a mbo-/mo- causative; lexical causatives are by definition exempt from this. It must be stressed that this type of derivational account certainly seems to be the right analysis for the Class I and II predicates reviewed above—both those that are aspectually polysemous or those that have only a change of state sense. As to the ‘feigned state’ sense, both analyses seem to account for the data discussed so far but, as will be seen, this does not hold when additional data are inspected.

Before looking at the relevant data, it is necessary to spell out the divergent predictions that the analyses make. First, the derivational account predicts the existence of the causative verbs—derived or lexical—that serve as the input of je-/ñe- derivation. The constructional account, in contrast, does not make such a prediction because, strictly speaking, no causative verb is involved in its schema; m(b)o- is not the true causative prefix but just part of the string -ñem(b)o. Second, the derivational account has it that the ‘feigned sense’ is only one of the senses some je-/ñe-derived predicates have along with reflexive and passive readings, which have a more central status. This predicts that a predicate that licenses the ‘feigned state’ sense will also convey reflexive and passive readings. The constructional account, for its part, dictates that the semantic component of the construction only has the ‘feigned state’ sense. In many cases, the instantiation of the construction and the output of the derivation may coincide superficially, but this is not necessarily so. Thus, this account predicts that the ‘feigned state’ sense might be formally dissociated from the reflexive and passive senses.

To begin with, consider the following examples.

(56)	a.	-ñemboguata	‘to pretend to walk in a certain way’/*‘to be made to walk’
	b.	-ñembopuka	‘to pretend to laugh’/*’to be made to laugh’
	c.	-ñemyasẽ	‘to pretend to cry’/*’to be made to cry’
			(Velázquez-Castillo 2002, p. 516)

The causatives of these predicates are acceptable and perfectly compositional: -moguata ‘make walk,’ -mbopuka ‘make laugh,’ and -myasẽ ‘make cry.’13 However, from this point on, the derivational analysis shows a problem. Recall one of the predictions that this analysis makes is that the ‘feigned state’ sense should co-occur with the reflexive or passive interpretation in the same form but, as can be seen in (56), the latter reading is not available to these predicates.14 The constructional analysis, conversely, deals with these examples in a non-problematic way. As the constructional schema indicates in (54), the slot for the verb has the ‘intransitive’ verb specification, making it possible for verbs such as -guata ‘walk,’ -puka ‘laugh,’ and -rasẽ ‘cry’ to fill it. Furthermore, since the construction is independent of the derived predicates, the formal dissociation of the ‘feigned state’ meanings and the passive and reflexive meanings is something to be expected, as is attested in (56).

A more complex scenario is posed by the following example.

(57)	A-ñe-mo-mano.
	1ac-agd-caus-die
	‘I feign being dead.’	(Estigarribia 2020, p. 286)

This presents challenges to the derivational analysis in terms of the two predictions mentioned above. First, the derivation in (55) indicates that the causative verb for this predicate should be -momano. Recall, however, that -mano ‘die’ is one of the few verbs that has a suppletive causative counterpart, -juka ‘kill’ (see Table 5). Thus, the derivation predicts the existence of a causative different from the one that is firmly established in the lexicon.15 However, even if -momano is perfectly acceptable, it still remains unanswered why the ‘feigned state’ sense should not be available to the je-/ñe-derived predicate that takes -juka ‘kill,’ illustrated in (58–59).

(58)	Fernando		o-je-juka.
	Fernando		3ac-agd-kill
	‘Fernando killed himself.’
(59)	O-je-juka	peteĩ		ovecha	ne-santo-ára-rehe.
	3ac-agd-kill	one		sheep	2sg.pos-sain-day-for
	‘A sheep was butchered for your birthday.’					(Velázquez-Castillo 2008, p. 393)

Note the derived predicate -jejuka has the reflexive (58) and passive (59) readings typical of je-/ñe-derived predicates. This is not possible for -ñemomano, which has only the ‘feigned state’ sense illustrated in (57).16 So, again, the derivational analysis fails to explain why the ‘feigned state’ sense is formally dissociated from the reflexive and passive senses. The constructional analysis, by contrast, handles these facts just fine. Since the constructional schema specifies that the verb filling the slot must be intransitive, -juka ‘kill’ is barred from appearing there, leaving -mano ‘die’ to take over for the expression of ‘pretend to be dead.’ -Juka is certainly not barred from serving as input to the derivational rule that yields the je-/ñe- form with the reflexive and passive interpretations in (58–59). Both the lexical rule derivation for the reflexive and passive readings and the constructional schema for the ‘feigned state’ sense readily explain the formal dissociation of the senses attested here and how they distribute across the two forms.

A final piece of evidence can be adduced against the derivational analysis and in favor of the constructional one. Some of the predicates in Class I can also give rise to a ‘feigned state’ interpretation. Predicates such as -jepicha ‘offended/take offence’ and -ñemondýi ‘scared/get scared’ can be associated with the general sense of the ‘feigned state’ interpretation (e.g., someone pretending to be offended). What is interesting about these predicates is that they are already derived by means of the je-/ñe rule, so under a derivational analysis, the prediction is that they should convey the ‘feigned state’ sense on their own. The constructional analysis, for its part, predicts that, being intransitive, these predicates can fill the slot for the verb in the schema and, thus, have the string -ñembo attached to them, yielding a form like -ñembojepicha. Consistent with all the previous evidence, this is precisely what the facts in the following examples reveal.

(60)	Context: We played a prank on Federico. It was really no big deal and he is a very easy-going person, but he wants to make us believe he is offended just to tease us.
		Federico	o-ñe-mbo-je-picha.
		Federico	3ac-agd-caus-agd-offend
		‘Federico pretends to be offended.’
	#	Federico ojepicha.
(61)	Context: Iván is reading his little nephew a story. To make it more entertaining, Iván acts out the reactions of the characters. When he gets to a scary part, he pretends to be very scared and his nephew is very amused by that.
		Iván		o-ñe-mo-ñe-mondýi.
		Iván		3ac-agd-caus-agd-scare
		‘Iván pretends to be scared.’
	#	Iván oñemondýi.

Just like the data in (57–59), these examples replicate the formal dissociation of the ‘feigned state’ interpretation from the reflexive and passive interpretation. Still, there is an additional problem for the derivational account here. Recall that one of the predictions of this analysis is that the causative form that serves as input to je-/ñe- derivation should exist. However, in the case of (60), speakers find *-mbojepicha unacceptable. This is not a problem for the constructional account as it does not predict the existence of such a form. With (61), the story is different because speakers accept -moñemondýi, which they take to mean ‘make/have someone scare one,’ as in amoñemondýi ‘I have someone scare me.’ This can be handled by the derivational account because nothing prevents the causative rule from taking a derived intransitive form such as -ñemondýi as its input. However, it is not obvious why the ‘feigned state’ sense should arise as a result of this second derivation and not the first one, that is, why this interpretation is obtained with -ñemoñemondýi and not -ñemondýi. There is no evidence to support such a second successive je-/ñe- derivation for the ‘feigned state’ sense in the other cases. This problem does not arise under the constructional analysis.

Finally, a fundamental point must be made against the derivational analysis and in favor of the constructional one regarding the semantic composition of the ‘feigned state’ sense. It is not at all clear how this sense would be obtained from je-/ñe- derivation. As was said in Section 3, the analysis of the prefix as an agent-demoting marker (Estigarribia 2020) readily explains all the eventive readings of je-/ñe-derived predicates from causatives. Thus, agent demotion might mean that the agent is left implicit or unspecified—as with passive readings—that it is totally suppressed—as with spontaneous occurrences—or that it is made to coincide with the patient—as with reflexives. In the case of Class I predicates, the stative sense is also compatible with the demotion of the agent as states are nonagentive. The ‘feigned state’ sense, however, does not seem to arise so easily from an agent-demoting derivation. All the data analyzed here point to the participant of a ‘feigned state’ as actually doing something purposefully, that is, as being unambiguously agentive. Note this is what is highlighted by the examples in (41) and (48) with the imperative (see also the comments in note 12). In keeping with this, ‘feigned state’ participants are invariably animate and typically human, as was shown in (42). In other words, ‘feigned states’ are agentive, so no agent demoting can be said to occur. Any attempt to preserve the derivational analysis would have to account for the exceptional behavior of je-/ñe- in this case. Under a constructional view, conversely, there is no need to reconcile the ‘feigned state’ sense with the senses resulting from je-/ñe- derivation.

In summary, it has been shown that the derivational analysis fails on three important counts. First, it wrongly predicts that the ‘feigned state’ sense should always be jointly expressed with the reflexive and passive senses, which is not supported by the evidence reviewed in (56–61). Second, it leaves the form in (60) unexplained as this cannot possibly result from the application of the lexical rule in (55). This is further compounded by the fact that the form in (61) suggests an otherwise unwarranted second derivation for the ‘feigned state’ sense. Third, the ‘feigned state’ sense entails an agentive participant so it cannot result from a derivation that demotes an agent. It follows, then, that the derivational analysis is unsuitable for the ‘feigned state’ interpretation. Conversely, the constructional analysis captures all the facts correctly. While the reflexive and passive senses co-occur with the ‘feigned state’ sense in the majority of the cases, this is an epiphenomenal coincidence stemming from the lexical rule and the constructional schema yielding superficially identical forms for most predicates. Additionally, the semantic idiosyncrasy of the ‘feigned state’ sense is explained by the independence of the construction. These unpredictable semantic and formal traits from the derivation are precisely the hallmarks of what an independent construction is according to Goldberg’s (1995) definition in (53). It can be concluded, then, that the ‘feigned state’ interpretation is the semantic component of an independent construction in Guarani. Most probably, this construction was calqued on the Spanish construction se hace el X, as in se hace el tonto ‘he pretends to be dumb’ or se hace el que llora ‘he pretends to be crying.’ More research is necessary to see if this hypothesis is correct.

6. Conclusions

This paper examined the distribution of state, change of state, and ‘feigned state’ senses across je-/ñe- predicates in Paraguayan Guarani. The question was considered against the backdrop of the broad extension of an aspectual polysemy pattern whereby many monovalent underived predicates have a state and change of state interpretation. Restricting the scope of attention to je-/ñe- predicates derived from causative verbs, two relevant different classes of derived predicates were recognized: the first one, called Class I, comprising those predicates derived from causative bases lacking an underived intransitive counterpart; the second one, called Class II, including those predicates derived from causative bases with an underived intransitive counterpart. It was shown that the state and change of state polysemy pattern carries over only to Class I predicates, which is understandable insofar as these are the only monovalent predicates corresponding to their causatives. Class II predicates, by contrast, give rise only to eventive interpretations such as reflexives or passives. This means the organizing principle for the distribution of the aspectual polysemy pattern across je-/ñe-derived predicates is whether or not the causative base predicate has an underived intransitive counterpart.

An apparent exception to this distribution is the ‘feigned state’ interpretation that some of Class II predicates receive. However, it was demonstrated that this is not a stative sense but rather an activity. Furthermore, it was argued that ‘feigned state’ je-/ñe- predicates are not obtained derivationally but constitute an independent construction. When a wider range of data is inspected, a constructional analysis according to which the ‘feigned state’ sense is its semantic component manages to capture the facts more successfully than an analysis that posits that the ‘feigned state’ sense is derived on a par with reflexive and passive senses.