In Xavánte the standard negator is õ (J. Santos, 2008, p. 87), illustrated in (2), repeated below. It derives from Proto-Akuwẽ *kõ (Nikulin, 2020, p. 130).

Estevam (2011, pp. 270-271, 277) treats õ as one of two allomorphs, with õ for the ‘indicative’ – the standard negator – as in (2), and tõ for the ‘subjunctive’, used as the prohibitive in (5).

The allomorphy with a t- form could tempt one to see õ and tõ as related to the form tũ in the Paraná languages Laklãnõ or Kaingáng, to be discussed in the section on Paraná *tũ^K. While this relation cannot be excluded, the parallel is not convincing. First, there is nothing in the Paraná languages corresponding to indicative – subjunctive allomorphy. Second, the syntax of standard negation is different. The Xavante standard negator õ has to be followed by a di element, considered by Estevam (2011) and Beauchamp et al. (2017) to be an ‘expletive’ element.

In Xerénte the etymological counterpart to Xavánte õ is kõ. Xerénte kõ also has to be followed by a di element, to the extent even that the grammarian Sousa Filho (2007, p. 140) writes kõdi as one word and glosses the whole thing as neg.

Notwithstanding the similarity between Xavánte (2) and Xerénte (6), grammarians do not agree about the status of di. Let us start with Sousa Filho (2007) analysis. Sousa Filho (2007) distinguishes three non-negative uses of di. In one use it is called a ‘predicative morpheme’ and it is usually glossed as pred (Sousa Filho, 2007).

What unites these uses, we propose, is that the existence of something, like a canoe, a small quantity or somebody’s good heart, is presented as a predicate of something else, like the river, the food or the speaker. The term ‘predicative’ is justified, but so is the term ‘existential’. ‘Existential’ may even be better, for we think that ‘existential’ as applied to di implies ‘predicative’, but not the other way round.

In a second, non-negative use, di is claimed to form participles.

It would not be a productive process, and Sousa Filho (2007, p. 164) leaves it open as to whether this use is different from the first use, now called ‘predicative-stativizing’.

A third use is illustrated in (9).

Sousa Filho (2007, p. 226) does not analyze this use, hence the question mark in the glosses. One wonders whether it could be a kind of topic (or focus) structure – at least, originally – with ‘Supra’ as topic (or focus), suggesting something like ‘It is Supra that gave me eggs’, in which English too uses a stative copula.

Xerénte di thus has a variety of related but slightly different uses, viz. of a predicative, stativizing and existential nature. Its negative uses fit here too: Sousa Filho (2007, p. 140) calls it ‘predicative’, while Miranda (2015, p. 265) and Cotrim (2016, pp. 119-120) use ‘stative’, and we don’t see any objection to calling this use ‘existential’.

This analysis is mirrored by most analyses of Xavánte, but not by all. Thus McLeod and Mitchell (1977, pp. 72-76), followed by Pickering (2010, p. 58), call di/ti a stativity marker and they make the claim that it can be used to predicate existence, especially in the negative.

J. Santos (2008, p. 87) calls di a marker of stativity and existence, and Oliveira (2007) calls it ‘stative’. In Lachnitt (1987, p. 22)di is glossed as ser, estar, ter, haver ‘be, have’ and in Hall et al. (2004, p. 31) as ‘a word that indicates a state or a position’. Quintino (2020), finally, considers di to be a stative copula.

A divergent hypothesis is offered by Estevam (2011), followed by Beauchamp et al. (2017). For Estevam (2011)di is either an expletive or an impersonal subject pronoun. It is expletive in negative structures, and impersonal with stative verbs. The latter is illustrated in (11).

Estevam (2011) agrees, though, about the existential origin of di, with an explicit reference to McLeod and Mitchell (2003) (Estevam, 2011, pp. 73, 220, 278, 440, 488). She cites an example like (10), making the point that this pattern is typical for the Culuene dialect studied by McLeod and Mitchell (2003), but no longer found in the São Marcos dialect that she studied. She does not make it explicit what the São Marcos dialect uses instead, but there are examples with the verbs höimana ‘live, exist, stay’ (Estevam, 2011, p. 269) and robaba ‘be absent, be empty’ (Estevam, 2011, pp. 104, 462). Quintino (2020) mentions the word õneharé1. Estevam (2011, p. 411) further argues that even in the São Marcos dialect, there are di uses that can still be taken to mark existence.

We find the hypothesis that di is a pronoun implausible. First, we are not aware of any cross-linguistic parallel for a development of an existential verb into a pronoun. Second, this change would result in a postverbal position for a pronoun, a position that is otherwise impossible in this language (Oliveira, 2007, p. 192)2.

Claiming a link between standard and existential negation urges one to bring in the notion of the ‘Negative Existential Cycle’ (‘NEC’). There are a few different approaches in the literature (see Croft, 1991; Veselinova, 2013, 2014; Veselinova & Hamari, 2021; van der Auwera et al., 2021). The basic idea is that first a dedicated expression for the negation of the existence of an object is also used for standard negation, viz. by expressing the negation of the existence of a state of affairs, and that these two uses later split (and thus prepare the ground for this kind of change to happen again). (13) and (14) illustrate this with pseudo-English. ‘There is/are no’ is used for denying the existence of an object - (13a) and (13b) shows that in pseudo-English ‘The hunter is not chasing the bear’ is expressed as a denial of the existence of a state of affairs of the hunter chasing the bear. This may be preceded by a stage in which pseudo-English was just like English and expressed ‘The hunter is not chasing the bear’ as, to wit, ‘The hunter is not chasing the bear’. It is this version of the NEC that is schematized in (14) and we see that ‘there is no’ takes over from ‘not’ for the expression of standard negation. What happens then is that pseudo-English makes another exponent for the existential negation of an object, e.g. ‘is non-existent’ – (13c). In addition, pseudo-English may drop the semantic existential component in the expression of standard negation relative to a state of affairs – (13d). In the latter case the form may still reflect the existential original – we represent the new standard negator deriving from an earlier existential negator as ‘[there.is.no]’.

The schema makes it clear that the third stage is similar to the first stage in having two different negators. But the negators are formally different from the ones in the first stage, with this special feature that the new standard negator derives from the old existential one.

(15) shows the variant in which pseudo-English had no exponent of standard negation other than a negative existential – or for which we don’t know that it had one.

In Xerénte and the Culuene dialect of Xavánte standard negation is construed as the non-existence of a state of affairs. In the São Marcos dialect of Xavánte, however, the marker is hardly used for negative existence, and the latter is expressed with something else. In this dialect õ di still has stative, predicative uses, but no longer the subtype used for existence. (16) is a sketch of the NEC for the São Marcos dialect of Xavánte.

Interestingly, both Estevam (2011) and Quintino (2020) mention examples in which õ takes care of standard negation without di.

It makes sense to see bare õ as a further development of the standard negator õ di: the di part could be dropped because õ di would not express negative existence anymore, nor would di express existence. On the other hand, in Estevam (2011) the majority of the di-less õ negators (11 out of 13) occur in subordinate clauses. We know from the literature on negation (van der Auwera, 2010, pp. 83-84; van der Auwera & Krasnoukhova, 2019, p. 24) that older negators may survive longer in subordinate clauses. So from this perspective they might be seen as preserving an older di-less negator.

Estevam (2011) does not herself allude to the NEC, and neither does she in the co-authorship of Beauchamp et al. (2017). In fact the latter provide an alternative account. They combine the observation that di still has positive stative uses, as in (11), with the observation that õ can be found with a privative meaning (‘without’).

Examples like (19) are mentioned by Lachnitt (1988, p. 73) and Quintino (2020) as well. They are all lexemes, not phrases, and we can think of this õ use as a derivational element. More importantly, could this seemingly privative use mean that õ di originally meant ‘be without’? We do know that there is such a thing as a Privative Cycle, similar to the NEC, with a language developing a standard negator from a privative marker (Michael, 2014b) and that the Cycles are very similar (van der Auwera & Krasnoukhova, 2020, p. 109).

But we are doubtful. First, it is not because (19) can be paraphrased with ‘without’ giving ‘things without use’, that one can conclude that õ is a privative morpheme. ‘Not use things’, ‘non-existent-use things’ would do as well, and in fact the gloss in Beauchamp et al. (2017), following Estevam (2011, p. 305), is simply ‘neg’. Of course, the notions of privative and existential negation are closely related: when somebody/something is without somebody/something, then the latter somebody/something does not exist with respect to the first one. So it is no surprise that in Veselinova (2013) 95 language-sample the existential and privative negators are identical in ten languages. In those cases, one could think of the negators as a kind of labile operator, for the entity that does not exist is also the entity that the states of affairs is deprived of (‘is without’).

So we do not exclude that the Xavante õ was a labile existential-privative operator. Second, ‘be without’ forms a positive predication, which ascribes a negative property, and it is also a stative predicate. So one would expect the positive stative predicate ‘be without’ to be expressed the same way as, for example, ‘be thirsty’. In Xavánte ‘be thirsty’ is indeed taken as stative predicate – and, as one might expect, not as an adjective, for the language does not have adjectives – and this is expressed impersonally.

So ‘be without’ would have to be construed impersonally, too. Thus (2) should not be paraphrased as ‘we are going to be without mingling’ but rather as ‘it is without us going to mingle’ or ‘there is without us going to mingle’. But these paraphrases amount to the existential ‘there is no us going to mingle’. So we conclude that at the relevant stage a would-be privative construal boils down to negative existence.

Intriguingly, Estevam (2011) does have a use for the ‘privative’ gloss, as in (22).

The first thing to note is that the use in (22) is not predicative but copredicative. It forms a secondary predication, subordinate to a main predicate. In (22) the main predication says that they will suffer more and the secondary one says that this will happen without water. All the examples in Estevam (2011) are copredicative. We could also note that the privative marker is not õ, but ãna. This is interesting, but it does not damage the privative analysis of õ. The latter could simply be the old privative, kept in lexemes and in the standard negator, and which was replaced in the copredicative use.

It is also interesting that ãna is reported with something like a standard negation use. (23) is one of the two examples provided by Estevam (2011, p. 463)3.

Both examples are elliptical questions. The speaker in (23) may just have heard that his/her mother is going to be present at some ceremony. Something like this is the contextually understood main predication, which is accompanied by the secondary predication.

What (23) shows, we propose, is that when the main predication is clear from the context, as when it has just been mentioned, the secondary predication can have the effect of a main predication. The process resembles insubordination, i.e., the conventionalized main predication use of a subordinate predicate. This is a cross-linguistically widely distributed phenomenon (Evans & Watanabe, 2016), illustrated in (25) and (26) with Chilean Spanish and English.

But there is also a difference with insubordination. The ‘that’ structures like that of (25) occur in a limited set of contexts (like astonishment) and the ‘if only’ structures like in (26) occur with apodoses that allude to an unpleasant consequence. This constancy allows for conventionalization. With the main predication use of a secondary predication like in (23), however, there are no typical contexts and hence the chances for conventionalization are slim.

To conclude about Xavánte, even an example of a seemingly standard negation use of a privative negator as in (23) has not changed our opinion on the origin of the present-day Xavánte standard negator õ. It is unlikely to have arisen from an earlier privative use in the context of the promotion of a secondary predication to the status of a primary predication.

Let us, finally, briefly return to Xerénte. The Xerénte standard negator is kõdi, and we offer the same analysis as for Xavánte. What furthermore pleads for the NEC is that di still has an existential use, illustrated in (7). There seem to be two privative markers. One is knã, which, a reviewer points out, must be cognate to Xavánte ãna, illustrated in (23), with both Xavánte ãna and Xerénte knã as regular reflexes of *kə̃nə̃4.

The second one is tõ, documented by Miranda (2015, p. 267). It seems to be a derivational use, just like in Xavánte (18). We return to this element in the section on Paraná *tũ^K.

To conclude about Xavánte and Xerénte, it is likely that the Xavánte õ di and Xerénte kõdi negators have an existential origin, i.e., that they arose with a NEC. This conclusion, however, is not quite final yet. The analysis in the next section will give us a reason to return to at least Xerénte and bring in a privative hypothesis nevertheless, not as a replacement of the existential hypothesis, but as an addition.

A final remark: õ and kõ must be very old. Nikulin (2020, p. 130) reconstructs them to a Proto-Jê neg form, because of the possible Karajá cognate -kõ (Ribeiro, 2012, p. 63).

What Nikulin (2020, p. 515) reconstructs for Goyaz as *kêt occurs in a few forms and a few languages – see Table 1. One form is kêt – or with a clause-final echo vowel khêrê/kêrê in Kĩsêdjê and Kajkwakhrattxi. Apinajé has hkêt as well as (h)kêt=nẽ, and they function in different ways. Dynamic verbs referring to non-habitual actions take (h)kêt=nẽ, and dynamic verbs referring to habits, stative verbs, nominal predicates and existentials take (h)kêt (Oliveira, 2005, p. 390). In (29) we see its use with a non-habitual action.

The =nẽ part is a factive clitic (Oliveira, 2005, p. 390), which also occurs on its own.

Another complex form is Kajkwakhrattxi kêrê-re, with an originally emphatic -re element, glossed as ‘diminutive’ by Camargo (2015).

As we can see in Apinajé (29), Oliveira (2005, p. 251) takes the lexical verb that combines with the (h)kêt negator to be non-finite. So do Beauchamp et al. (2017) and Gildea and Castro Alves (2010, p. 187) and they all furthermore take (h)kêt to come from a negative existential (Oliveira, 2005, p. 390; Gildea & Castro Alves, 2010; Beauchamp et al., 2017). Miranda (2015, p. 215), however, tentatively proposes ‘not do’ as the original meaning. We will come back to the latter idea later. (32) and (33) illustrate existential uses.

In Apinajé existential (h)kêt alternates with hamrakati.

This is not a problem for thinking that standard (h)kêt derives from an existential. The NEC hypothesis allows a language that has construed standard negation with a negative existential to make a new existential negator.

In Mẽbêngôkre kêt also has a privative use – illustrated in (35), cf. also Trevisan & Pezzotti (1991, p. 43), whose lemma for kêt has ‘not, without, be nothing, be extinct’. It is reported as one of at least two privative markers in Krahô, the other one being =nõ5, but in this language kêt is used for neither standard nor existential negation.

So like for Xavánte õ, the *kêt negator has both existential and privative uses. But different from Xavánte scholarship, the linguists who have pronounced themselves on the origin of the standard negation use only adopted an existential analysis, not a Beauchamp et al. (2017). privative analysis. But there could be another factor advocating a privative approach.

This other factor is a process through which an extra-clausal left periphery topic (or focus) gets integrated into the clause and becomes its subject – described in the context of the ‘nominative-absolutive alignment’ hypothesis put forward by Castro Alves (2010), followed by Gildea and Castro Alves (2010, 2020). It is a process hypothesized for a variety of grammatical operators, not just negators. Extra-clausal left-periphery topic structures are considered a property of all the Goyaz languages. When the topic is a pronoun, it usually takes nominative marking and when it is a noun, it is unmarked. The front position constituent is coreferential with another constituent in the clause, whatever case marking this coreferential constituent has there. In (37) we first get the nominative pronoun ga, which is coreferential with a-je, which is ergative. The reason why the pronoun a-je is ergative is that subordinate clauses, including the non-finite constructions that go with the negator, have ergative-absolutive alignment. This topic then becomes integrated into the clause, it becomes its subject, and since there is also the coreferential subject argument of the erstwhile subordinate clause, we can speak of ‘subject doubling’ (cf. Salanova, 2007, pp. 34-35)6.

In some structures in some languages, the integration is very strong. It can exert influence into the subordinate clause by forbidding it to contain an ergative constituent and forcing coreferentiality with the absolutive argument – hence the label ‘nominative-absolutive alignment’. We see this in (38). When the clause-initial constituent is a noun, there cannot be a coreferential ergative pronoun.

If this analysis is correct, then, we propose, the negator changes its status. Earlier it was an intransitive non-existence verb with the non-finite verb form as its subject. Now, it has the nominative clause-initial constituent as its subject and it becomes quasi-transitive, with the non-finite verbal constituent as its object. It could be seen as a ‘not do’ auxiliary, picking up the suggestion by Miranda (2015, p. 215). However, it is more plausible to take it as a privative auxiliary, given the fact that kêt has privative uses and given that the step from ‘not exist’ to ‘be without’ is smaller than the one between ‘not exist’ and ‘not do’ – see the discussion around (20). Schematically:

We thus end up embracing both the existential and privative hypotheses on the origin of *kêt. We claim that *kêt is originally and in some constructions still an existential negator, and that the introduction of a topic-derived subject turns the existential into a privative. (41) adds the path from the negative existential to the privative onto the general schema in (15). Note that ‘there is/are no’ keeps the form of the existential negation – at least initially – but that its meaning is privative.

The reanalysis of the topic as a subject like the reanalysis of the negator is a gradual process. We have seen that in Kĩsêdjê the ban on the ergative inside the subordinate structure and the reanalysis of the negator only works with nouns (see (38)). It does not work with pronouns. The latter do not normally (cf. Gildea & Castro Alves, 2010, pp. 188-189) allow the topic construction, and there is no reason to deny that standard negator in (42) is anything other than an existential negator.

In Apinajé, however, the integration applies to pronouns too (Gildea & Castro Alves, 2010, pp. 180-181).

We will see in the section on Timbira *inõare that Canela has gone through the process of ‘privativization’ of the existential negator, too, but with a different negator. Canela, like the *kêt languages, is a Goyaz language. Could this process be unique to Goyaz? Gildea and Castro Alves (2020, p. 90) remark that what we call ‘clefting’ is indeed found throughout Goyaz, but that does not mean that it cannot occur elsewhere. In fact, we have already suggested that we may see clefting in Xerénte (9), where it could be argued to have the stativizer di. It is also useful to review the Xernte example (6), repeated below.

There are two exponents for the third person actor, a nominative and an ergative one. To the extent that the nominative one is integrated in the clause, a privative reanalysis of the existential negator seems possible.

A final word on the proto history of *kêt. Nikulin (2020, p. 441) reconstructs *kêt to Proto-Jê and there are reflexes in Paraná. It does not seem to occur in the wider Macro-Jê family. If the kêt forms originally expressed negative existence, then there are two possibilities for deriving this meaning from something else (but cp. Oliveira, 2005, p. 298). One possibility is that the original kêt word is a univerbation of a negator and an existence marker. One wonders whether kêt could derive from the ancestors of Xerénte kõ and di. The alternative etymology would derive kêt from a word with a negative meaning. Curiously, Nikulin (2020, p. 515) lists a Canela form -hkêt with the meaning ‘stop’ as ultimately deriving from a Proto-Jê negator8. Could the modern ‘stop’ meaning be a remnant of an ancient ‘stop’ meaning? In any case, the next section will show that the Jê languages do manifest trajectories with negators deriving from something close to a ‘stop’ verb.

An indication that a standard negator can develop from a verb meaning ‘finish’ comes from Jolkesky (2010, pp. 215, 244), who reconstructs *tũ(-ɡ) for Proto-Paraná Jê, both for a standard negator and for a lexical verb meaning ‘complete, not have anymore, finish (something)’. Jolkesky (2010), however, does not connect the two etymologies. Nikulin (2020, p. 430) discusses only the etymology of the negative morpheme, he reconstructs it to Proto-Jê *tũ^K with ‘negation’ as the meaning, he does not discuss the ‘finish’ verb, but he thought that it was obvious that the ‘finish’ and negative meanings are related (A. Nikulin, personal communication). Neither Jolkesky (2010) nor Nikulin (2020) express any view on whether the *tũ(-ɡ) form could express existential negation. When we look at the present Paraná forms deriving from the Proto-Paraná Jê *tũ(-ɡ), nobody has suggested a link either between verbs of ‘finishing’ and standard negation or between ‘finishing’ and existential negation. Nevertheless, the identity is clear in at least present-day Kaingáng, as described by Wiesemann (2011, p. 89). As to a link between existential and standard negation, Miranda (2015, p. 272) lists tũ for both Kaingáng standard and existential negation, but this is maybe only based on a dictionary lemma including ‘inexistente’ in the lemma for tũ (Wiesemann, 2011, p. 89). M. S. Silva (2011, pp. 156-157, 173) has examples. The meaning sem ‘without’ is in the Wiesemann (1972, 2011) lemma too, so perhaps there is a standard – privative polyfunctionality as well (cf. also Gonçalves, 2007, p. 192). This polyfunctionality is definitely also found in Laklãnõ (Gakran, 2015, pp. 204, 206), as illustrated in (44)9.

These (potential) polyfunctionalities show the affinity of standard negation to both existential and privative negation, but they don’t tell us anything about the diachrony.

As mentioned in the introduction to this section, the general typology prepares us to see an especially close link between a ‘stop’ verb and prohibitive negation. There is no evidence for *tũ^K ‘finish’ to have a special affinity with the prohibitive. Thus Laklãnõ uses a *tũ^K form for both standard and prohibitive negation (Gakran, 2015, pp. 203-208), and Kaingáng uses the *tũ^K form only for standard negation. Compare (44a) with (45) for Laklãnõ and (3) with (46) for Icatu Kaingáng.

It is to be noted that Gakran analyzes the -g element (phonologically /-ŋ/) as a causative suffix. The velar nasal is absent in (44), but the distinction is not simply between standard and prohibitive negation. According to Gakran (2015, p. 207) in standard negation tũ-g is used for perfectives and tũ for imperfectives. We have nothing to contribute on this issue (cp. also note 150 in Nikulin, 2020, pp. 293-294).

It is clear, however, that tũ is an old form. The form that Nikulin (2020, pp. 430, 454) gives for Proto-Jê is also the origin of the tõ negator found in various functions in the Goyaz languages Xavánte, Xerénte and possibly11 Panará. For Panará, Nikulin (2020, p. 454) also claims that the reflection of the protoform would be rõ, and this is used for the privative (Dourado, 2001, p. 118). There are also cognates in the wider Macro-Jê family, both in the Trans-São Francisco and Jabutian groups (Nikulin, 2020, pp. 118, 387).

To conclude: *tũ^Kis widespread and old. An origin in a ‘finish’ verb is not excluded, but the evidence is not very strong. There is no evidence in favor of a Negative Existential or Privative Circle either. In fact, as the next section will show, if *tũ^K originated from a ‘finish’ verb, it is likely that neither hypothesis is correct, i.e., that it followed a somewhat different path.

Dourado (2001, p. 119) states that the Panará standard negator pjo is related to an intransitive verb pjo/pjoo ‘finish’.

That Dourado’s (2001) phrase ‘related to’ means ‘derives from’ is plausible from the fact this verb also occurs with the ‘finish’ sense bleached into negation. This would then represent an intermediate stage in the development of a ‘finish’ verb to standard negation.

If Dourado (2001) is right and the meaning of the pjo verb in (48) has bleached, one can say that it expresses non-existence: when a process or event is finished, it does not exist anymore. The discontinuation sense is still present in (48) but perhaps the verb can also be used to refer to the non-existence of a state of affairs, independently of whether a process or event is finished or has not even started. If that is possible, then the verb has widened its sense: from non-existence due to finishing to non-existence tout court. This scenario is schematized in (49).

This notion of existential negation is, of course, a key ingredient in the NEC, but it plays a different role there. In the NEC a negative existence sense applies to both objects and states of affairs. The NEC has a stage in which existence is denied to a state of affairs in the same way as it is denied to an object. There is no such stage in (49). What we see therefore in Panará is a scenario that is similar to the NEC, but it is crucially different. This scenario is also different from what van Gelderen (2021, p. 545) sketched in (50).

In (50) we see a negative verb feeding into the NEC. It is not to be denied that this is possible. But what we see in Panará is different: in Panará the verb skips the second stage of (50) and for objects the existential negator is different, viz. inkjoo, shown in (51).

Panará does not use pjo as a privative negator either. Here rõ is used. We will come back to rõ at the end of the next section. Pjo is also not used for prohibitive negation, as becomes obvious from (52).

To conclude, the Panará pjo is only used for standard negation. We hypothesize that it comes directly from the ‘finish’ verb and that it neither passed via an existential negation ascribed to an object nor via a privative negation. In synchrony, pjo is not used for existential or privative negation either. Given the typological literature one might have expected it to serve as a prohibitive negator but that it is also not the case. The reason might be that in the examples known from the literature, the ‘stop’ verb is a transitive one, with an actor subject, which in a prohibitive context, becomes the addressee that the prohibitive appeals to for action. For Panará, however, Dourado (2001, p. 119) makes it explicit that the source verb is intransitive, so one that involved something finishing, rather than somebody finishing something. This intransitive verb is therefore not immediately suitable for an appeal to an action. And we therefore do not find evidence for a close link between the Panará pjo and prohibitive negation.

In awareness of the grammaticalization literature deriving standard negation from negative verbs, Castro Alves (2010, pp. 468-469) claims that the Timbíra standard negator nare (from *inõare, A. Nikulin, personal communication, 2020) is a grammaticalized form of ‘finish’12. A standard negation example is shown in (53). (54) illustrates a full verb use in closely related Mẽbêngôkre (cf. also Costa, 2015, pp. 189-192).

Next to the nare form illustrated in (53), there is also the short form na (Castro Alves, 2004, p, 129). This kind of variability fits a grammaticalization hypothesis.

Like for Panará we should ask how the Timbíra ‘finish’ verb acquired a standard negation sense. Castro Alves (2010, p. 468), followed by Gildea and Castro Alves (2020, pp. 87-88), explicitly says that the change involved an intermediate existential negation sense. This claim does not say what kind of existential negator it is. Like for Panará pjo we are not aware of any evidence that nare became an existential operator for objects. Of course, we don’t know about the earlier stages, but at least in present-day Krahô we find a negative verb jamrẽare as an existential negator (55).

There is also no present-day evidence for privative or prohibitive uses. As mentioned already, privative negation is served by either kêt, illustrated in (36), or nõ, and the latter also appears with a prohibitive function.

Nare thus patterns very much like pjo. Both demonstrably come from ‘finish’ verbs, they are currently only used in standard negation, not existential, privative or prohibitive negation. We may assume, in agreement with Gildea and Castro Alves (2020, p. 87), that the ‘finish’ verb that is the source is intransitive.

Yet there are two complications. First, nare, in (57) showing up in a short from na, has developed nominative-absolutive alignment.

The starting point is again, we propose, like for kêt, the integration of a topic/focus constituent integrating into the clause and then turning the existential into a privative, from ‘as far as you are concerned, there is no grabbing it’ to ‘you are without grabbing it’. We also see that the ‘privativization’ applies to pronouns, like it does for Apinajé, but not for Kĩsêdjê (see section on Goyaz *kêt). So to this extent at least Canela does have a privative use of nare and in this respect nare is like kêt. There is no such evidence for pjo. Panará alignment is altogether different, with a generalized ergative-absolutive system.

A second complication concerns the Panará negator rõ. According to Gildea and Castro Alves (2010, p. 87) it is a cognate of nare. Their source, Dourado (2001), does not actually say this, and Nikulin (2020, p. 454) suggests that rõ might derive from *tõ. In any case, its current patterning is not, in any strong way, related to that of nare or tõ. Rõ differs from tõ in that tõ is Paraná phenomenon and rõ is not. Rõ also differs from nare in that rõ has a wider spread. In Panará, rõ is a standard negator as well as a privative one.

We conclude that the question of the origin of rõ and its relation to nare remains open.

Autor para correspondência: Johan van der Auwera. University of Antwerp. Prinsstraat 13, Antwerp 2000, Bélgica (johan.vanderauwera@uantwerpen.be).