SciELO - Scientific Electronic Library Online

vol.13 issue2Dative subject in CanelaCarnival at the museum and Mamãe’s adventures: reflections on museum objects, cultural performances, and urban space author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand




Related links


Boletim do Museu Paraense Emílio Goeldi. Ciências Humanas

Print version ISSN 1981-8122On-line version ISSN 2178-2547

Bol. Mus. Para. Emílio Goeldi. Ciênc. hum. vol.13 no.2 Belém May/Aug. 2018 


The historical phonology of Paunaka (Arawakan)

Fonologia histórica do Paunaka (Aruaque)

Fernando O. de Carvalho1

1Universidade Federal do Amapá. Macapá, Amapá, Brasil


This paper applies the comparative method to unravel the historical development of the segmental phonology of Paunaka, an Arawakan language of Bolivia. Although the Paunaka vowel system features a single back rounded vowel, it is rather simple to show that it derives from a system with two back rounded qualities *u and *o, but that the former segment shifted to a high central unrounded vowel ɨ. The language has lost *r unconditionally, implying that Paunaka items with r are probable loanwords. Paunaka underewent a spirantization of *ts, thus merging this affricate with the fricative *s. Although Paunaka shares a coronalization of *k > s with Proto-Mojeño, most of the phonological developments that affected Paunaka are either recurrent in the Arawakan language family or only superficially similar to developments in related languages, and thus provide little weight as evidence for subgrouping. An Appendix is also included, with 105 etymologies matching Paunaka lexical and grammatical morphemes with their cognates in Proto-Mojeño, the two extant Mojeño dialects (Ignaciano and Trinitario) and Terena.

Keywords Comparative reconstruction; Sound change; Arawakan languages; Paunaka


Este trabalho emprega o método comparativo com o objetivo de elucidar o desenvolvimento histórico da fonologia segmental do Paunaka, uma língua Aruaque da Bolívia. Embora o inventário vocálico do Paunaka tenha apenas uma vogal posterior arredondada, é relativamente simples mostrar que o mesmo se deriva de um inventário com duas qualidades vocálicas posteriores arredondadas, *u e *o, mas que o primeiro desses segmentos se tornou uma vogal central não arredondada ɨ. O Paunaka perdeu *r em todos os contextos, um fato que em si sugere que formas contendo r sejam empréstimos com entrada na língua em data posterior a essa mudança. O Paunaka foi sujeito a fricativização de *ts, que, assim, têm reflexos idênticos aos de *s. Embora o Paunaka e o Proto-Mojeño apresentem a coronalização *k > s, a maior parte dos desenvolvimentos fonológicos do Paunaka são recorrentes em diversas línguas da família ou possuem apenas uma similaridade superficial com desenvolvimentos ocorridos em outras línguas, não sendo, portanto, muito relevantes para o estabelecimento de uma classificação interna. Por fim, é apresentado no Apêndice um conjunto de etimologias contendo cognatos no Proto-Mojeño e no Terena de 105 morfemas lexicais e gramaticais do Paunaka.

Palavras-chave Reconstrução comparativa; Mudança sonora; Línguas Aruaques; Paunaka


Paunaka (ISO 639-3: pnk) is a severely endangered Arawakan1 language spoken by less than ten individuals near Santa Cruz de la Sierra, Bolivia (Danielsen; Terhart, 2014). Until recently, all the evidence available on this language consisted of a couple of poorly-transcribed wordlists (Cardús, 1886, p. 319). Fortunately, however, the language has been under documentation by Lena Terhart and Swintha Danielsen, and much more extensive and reliable data on this language has become available in the last few years. Because of this paucity of data, Paunaka has not figured in most comparative investigations of Arawakan languages, apart from a recent study by Jolkesky (2016).

This paper examines comparative evidence to throw further light on the historical development of Paunaka phonology while at the same time addressing issues of broader significance for Arawakan historical linguistics. One central issue in the phonological reconstruction of Proto-Arawakan is the uncertain status of the contrast between two back rounded vowels (Payne, 1991, p. 476-478; Aikhenvald, 1999, p. 75-78). As Paunaka is one of the many languages in the family lacking this contrast (see the section ‘An outline of Paunaka phonology’), its inclusion in comparative investigations could provide additional evidence for the absence of the opposition between o and u in the proto-language. However, it is straightforward to show that the unrounded vowel ɨ, which, in this language, contrasts with rounded u, is an unconditioned reflex of *u, while u derives from *o, thus tracing the Paunaka system to an inventory which, just like Proto-Mojeño and Terena, does contrast two back rounded vowels (see the section ‘Vowel inventory’). For consonants, Paunaka cognates are crucial for showing that a fricative debuccalization change occurred independently in both Terena and Mojeño (see the section on ‘Consonant correspondences’). Concerning the implications of sound change developments for internal classification, vowel correspondences suggest that both Paunaka and Baure show the fronting of a back rounded vowel and that this could potentially constitute a (so far unrecognized) shared innovation. As discussed here, though, the developments inferred for the phonological history of Paunaka offer very little to the problem of internal classification, being either exclusively attested in this language or only superficially similar to developments attested elsewhere. However, before presenting the main developments in the historical phonology of Paunaka, a brief discussion of the internal classification of the language and the sample of languages chosen here for comparison with Paunaka are the topic of the next section.


The languages usually taken to be particularly closely related to Paunaka are: Mojeño, with its two extant varieties Ignaciano (ISO 639-3: ign) and Trinitario (ISO 639-3: trn), Baure (ISO 639-3: brg), Paikoneka (no ISO code; Glottocode: paic1240) and Terena (ISO 639-3: ter) (Aikhenvald, 1999, p. 67; Campbell, 1997, p. 181, 2012, p. 75; Danielsen, 2011, p. 517; Danielsen; Terhart, 2014, p. 221). I will focus here on a comparison with Terena and Mojeño, as inspection of cognate sets has suggested that these languages are the most informative as far as Paunaka historical phonology is concerned (as shown in the Appendix, the proportion of shared cognates between Paunaka and Mojeño is particularly impressive). Focusing on a comparison with Mojeño and Terena seems justified for the following reasons: first, Paikoneka can be excluded without loss since it is already extinct and only very superficially attested (Danielsen; Terhart, 2014, p. 222). Second, Modern Baure, the best attested variety of this language, seems to be exceedingly innovative in its phonology, so much that cognate identification is many times hampered by vowel losses (mostly apocope) and changes in vowel quality (umlaut). This is illustrated in Table 1 by a comparison of forms from sources on Old Baure (18th century; see Adam; Leclerc, 1880) and their correspondents in Modern Baure (data from Danielsen, 2013, p. 288, I retain the source orthography; see Danielsen, 2007, p. 51-55 for further discussion)2.

Table 1 Comparison of Old Baure and Modern Baure forms. 

Old Baure Modern Baure
Woman eteno eton
Hand wejise -wojis
Ear chakane -chokon
Moon kejere kijer

Some amount of systematic internal reconstruction comparing different Baure dialects and making full use of the documentation available on Old Baure is necessary before the language can be profitably used for casting light on the development of an apparently more conservative related language (I will come back to Baure, however, in the final section of this paper, where the issue of internal classification will be discussed). A third factor justifying this narrowed comparative focus relates to the extent of the available documentation on Paunaka and to the immediate goals of the present study. Given that our goal of elucidating the historical phonology of Paunaka is necessarily founded on establishing a sizeable number of etymologies matching Paunaka forms and their cognates in other languages, restricting the comparison to particularly closely related languages will certainly yield a higher number of cognates, specially so if the languages brought into comparison happen to be particularly well-documented. Mojeño and Terena fit both desiderata: first, preliminary evidence suggests that Paunaka is particularly closely related to Mojeño (Danielsen; Terhart, 2014; Jolkesky, 2016) and, in addition, that Terena and Mojeño are also closely related (Carvalho, 2017a). On the availability of extensive data, for Mojeño there are three dictionaries: Gill (1993) on the Trinitario variety, Marbán (1701) on 17th century Old Mojeño and Ott, W. and Ott, R. (1983) on the Ignaciano variety. These are complemented by a grammar of the Trinitario variety (Gill, 1957), a stream of recent in-depth descriptions of parts of Trinitario phonology, morphology and syntax in Rose (2011, 2014, 2015a), an extensive description of the morphosyntax of Ignaciano (Olza Zubiri et al., 2002) and the grammatical description of Old Mojeño in Marbán (1701). I have also benefited from the Proto-Mojeño (PM) reconstructions of Carvalho and Rose (2018) for comparison with Paunaka. For Terena there are several papers and book-length descriptions authored by Summer Institute of Linguistics (SIL) linguists (Ekdahl; Grimes, 1964; Eastlack, 1968; Bendor-Samuel, 1961), an extremely useful pedagogical grammar in two volumes (Ekdahl; Butler, 1979), an unpublished dictionary (Ekdahl; Butler, 1969) and first-hand, fieldwork data by the author of the present paper. For Paunaka, I rely on Danielsen and Terhart (2014) and Terhart (2014), the former mainly but not exclusively for a description of the phonology and morphology of the language and the latter as a privileged source of lexical data.


The Paunaka inventory of contrastive vowels and consonants is given in Tables 2 and 3, after Danielsen and Terhart (2014).

Table 2 Paunaka vowels. 

[- Round] [+ Round]
Front Central Back
High i ɨ u
Mid e
Low a

Table 3 Paunaka consonants. 

Labial Alveolar Palatal Velar Glottal
Oral Stop p t k
Affricate ʧ
Fricative ß s h
Nasal stop m n
Liquid r
Approximant j

In its vowel system, Paunaka differs from both Proto-Mojeño and Terena in lacking a contrast between two back rounded vowels and having a contrastive high central vowel ɨ. As discussed in ‘Vowel inventory’ though, the Paunaka vowel system is clearly derived from a system with two contrastive back rounded qualities3.

The set of contrastive consonants in Table 3 is based on Danielsen and Terhart (2014) but does not include non-contrastive segments or those whose distribution is restricted to obvious loanwords, such as the glottal stop ʔ and the palatal fricative ʃ.

Paunaka, just like Proto-Mojeño, has a single liquid consonant, a rhotic. Synchronic instances of r in the language likely result from borrowing, as comparative data shows that the language has lost *r regularly. The language differs from Proto-Mojeño and from Terena in lacking a contrastive nasal stop ɲ, which occurs only as a surface variant in Paunaka (Danielsen; Terhart, 2014), and, also, by lacking an alveolar affricate ts, reconstructed for Proto-Mojeño and inferable for earlier stages of Terena (Carvalho; Rose, 2018; Carvalho, 2017c). Finally, like most Arawakan languages, Paunaka organizes its segmental elements in simple CV syllables, eventually tolerating onset-less V syllables in word-initial position and allowing for some amount of tautosyllabic vowel combinations (Danielsen; Terhart, 2014, p. 229). Examples of allowed vowel sequences include: ai, in kupisaire ‘fox’; ae, in the Locative marker jae (though here ae may be a contextual variant of ai4); au, in nauku ‘there’; ue, in kuepi ‘potato’; ui, in nisuika ‘I will write’; ei, in tuseina ‘noon’ and ɨu, in the Demonstrative ʧɨu. Many of these vowel sequences are found exclusively or frequently across morpheme boundaries only, such as iu, in niuma ‘my grandfather’. As seen in the next section, many of the vowel clusters result from the historical loss of an intervocalic consonant, either the rhotic *r or the glottal stop .


The segments of the Paunaka inventory appearing in Table 4 have a simple history, being either retentions or the result of phonetic shifts that did not affect previously existing contrasts. Note as well that a glottal stop was unconditionally lost in the language. Both the cognates for each supporting set and the segments in each correspondence are here and elsewhere presented in the order Paunaka, Proto-Mojeño (PM), Terena. For each cognate set a single meaning gloss is given, arguably approximating that of the etymon. The full list of cognate sets given as an Appendix to this paper includes commentaries and notes for those cases where the meaning of one or more cognates deviates from that given in the semantic entry.

Table 4 Correspondence sets. 

Proto-segment Paunaka PM Terena
(a) *p p *p p
(b) *m m *m m
(c) *w ß *w w
(d) *j j *j j
(e) *n n *n n
(f) ʔ
(g) *a a *a a
(h) *e e *e e
(i) *i i *i i

Supporting cognate sets for each of the correspondences in Table 4 are given below:

(1) Exemplar cognate sets for each diachronic correspondence

p < *p : Wash -kipu : *-sipo : -kîpo, Fear -piku : *-piko : -pîko,
Bone -upe : *-ope : -ôpe.
m < *m : Earth mute : *móte-hi : móte, Husband -ima : *-ima : -îma,
Tapir samu : *samo : kámo.
ß < *w : Be, Stay -ußu : *-owo : -ôwo, Take -ße-u : *-weʔo : -wêo,
Foot -ißu : *-iwo-pe.
j < *j : Wife -jenu : *-jeno : -jêno, Night juti : *joti : jóti, Hair -hiju : -hijo.
n < *n : Tongue -pe-nene : *-nene : -nêne, Go -junu : *-jono: -jôno,
Jaguar isini : iʧini : sîni.
< *ʔ : Hand -ßuɨ : *-woʔu : -wôʔu, Hit -eu : *-eʔo,
Soil apuke : *apókeʔe : pokéʔe.
a < *a : Sun saʧe : *saʧe : káʃe, Ear -ʧuka: *-ʧoka, Worm kane : *kane : kâne.
e < *e : Breast ene: *-ʧene: êne, Back -kekɨ : *-keku : -keku,
Pet -peu, *-pero, -pêjo.
i < *i : Husband -ima : *-ima : -îma, Stone mai: *mari: marîpa,
Jaguar isini: *iʧini: sîni.

On the postulated phonetic shift *w > ß, note that Paunaka, Mojeño and Terena have a single contrastive consonant whose realization fluctuates between [w] and [ß] (Danielsen; Terhart, 2014, p. 228). Phonological arguments, related to nasality spreading, can be offered for analyzing the Terena consonant in question as underlyingly sonorant, hence w (Carvalho, 2017b). Though w and ß seem to have become marginally contrastive in the Trinitario dialect of Mojeño, reconstruction of PM *w is uncontroversial (Carvalho; Rose, 2018). For these reasons, and since a similar pattern is attested far afield in the Arawakan family5, I have reconstructed *w for the common ancestor of Paunaka, Mojeño and Terena, thus implying the minor phonetic shift *w > ß6.


Paunaka differs from both Mojeño and Terena in having a single back rounded vowel, u, while both Proto-Mojeño and Terena show u and o. This fact could suggest a simple merger of the two back rounded vowels in Paunaka, a simplification similar to the merger of PM *a and *o that took place in the Ignaciano variety of Mojeño (Jolkesky, 2016; Carvalho, 2017a; Carvalho; Rose, 2018)7. Comparative data shows, however, that the two non-front, non-low vowels of Paunaka, u and ɨ, reflect two back rounded vowels, here reconstructed as *o and *u respectively (see Table 5).

Table 5 Correspondence sets. 

Proto-segment Paunaka PM Terena
(a) *o u *o o
(b) *u ɨ *u u

(2) Exemplar cognate sets for each correspondence
(2a) Paunaka u: PM *o: Terena o
Wing/shoulder -pußu: PM *-powo: -pôwo, Moon kuhe : *kohe : kohêe, Night juti: * joti: jóti,
Woman esenu: *eseno: sêno.
(2b) Paunaka ɨ: PM *u : Terena u
Hand -ßuɨ: *-woʔu: -wôʔu, Cloud ɨku : *uko : úko, Firewood -jɨkɨ-ke : *-juku-ki : -júku, Ant kusiɨ: *koʧiru: kosîu.

Jolkesky (2016, p. 17) postulates a third correspondence according to which Paunaka u would match Mojeño u as well, thus overlapping with both (2a) and (2b) above (he did not include Terena in his comparative study). This has led Jolkesky (2016) to assign correspondence (2b) above to , since *u was already reconstructed for this third, identity correspondence. Accordingly, he postulated a merger of two back rounded vowels *u and *o in the history of Paunaka and a merger of and *u in Mojeño (Jolkesky, 2016, p. 30). A critical problem with this proposal is that this putative identity correspondence reflecting *u is an artifact and, actually, does not exist8. A look at his comparative data (Jolkesky, 2016, p. 18-24) reveals only two forms for which the correspondence Paunaka u : PM *u is attested, and both are problematic. For the existing, attested Paunaka noun ɨʧɨ ‘capybara’, Jolkesky (2016, p. 19) presents instead a ‘Pre-Paunaka’ form *uʧɨ for which no motivation or evidence exists. Comparing instead the attested Paunaka form ɨʧɨ with PM *uʧu ‘capybara’ (see Appendix) yields an instance of correspondence (2b) above. The other etymology which apparently supports this third correspondence is PM *koʧiru : Paunaka kusiu ‘ant’. My inspection of first-hand data on Paunaka from Lena Terhart’s work has revealed, however, a different form for ‘ant’, kusiɨ, which is completely consistent with correspondence (2b) above (and with other regular correspondences as well), here reconstructed as a reflex of *u and implying thus a change *u > ɨ for Paunaka.


The following sections address the most difficult problems in Paunaka historical phonology: the development of the coronal obstruents, the reflexes of *s and *h, the status of the rhotic r and the correspondences involving the velar stop k.


Below we have correspondence sets for alveolar fricatives and affricates, except those that are clearly reflexes of *k (these will be discussed in the section ‘The velar stop and diachronic fronting’ since they overlap in a crucial way with the identity correspondence for the velar stop *k).

Correspondences (a), (b) and (c) are here reconstructed as reflexes of *ts. De-affrication of ts to s is independently supported for Terena9. The same development can be postulated for Paunaka, which is consistent with the fact that the language lacks ts altogether (Danielsen; Terhart, 2014, p. 227). Though both Terena and Paunaka agree in showing the same reflex for all correspondences (a-c) in Table 6, an alveolar fricative s, PM calls for the establishment of the three correspondences attested and the varying PM reflexes must be accounted for somehow. The supporting evidence for each correspondence is given in (3a-3c) below, followed by discussion.

Table 6 Correspondence sets. 

Proto-segment Paunaka PM Terena Contexts
(a) *ts s *ts s ___ *e, *i
(b) *ts s s (*i) ___*i
(c) *ts s *s s ___ *o, *a, *e
(d) ʧ t ___ *u
(e) ʧ ʃ ___ *o, *e, *u
(f) *t t *t t ___*e, *i

(3) Correspondences and exemplar cognate sets for *ts
(3a) Paunaka s : PM *ts: Terena s /__ *e, *i
Grandmother -use : *-otse : -ôse, Ashes sima-pa : *tsima-pa,
Red tisi : *titsi, Eyelash musipa: *motsi-pa.
(3b) Paunaka s : PM *ʧ : Terena s / (*i)__*i
Jaguar isini : *iʧini: sîni, Ant kosijɨ : *koʧiru : kosîu.
(3c) Paunaka s : PM *s : Terena s /_ *o, *a, *e
Weed out -su : *-iso-ʔo : -íso, Garden -asane-ti : *esane-ti : isáne,
Defecate -suku : *-soko, Woman esenu : *eseno : sêno,
Mother-in-law -muse : *-ímose : -imóse.

Dealing first with the Mojeño reflexes, if PM *-otse ‘grandmother’ is ignored for a moment, the correspondence in (3c) can be collapsed with (3a) and (3b), as the latter two are restricted to the context of a following *i. A single proto-segment, *ts, can be assumed for all three, *ts > s being a development restricted, in PM, to the context /__ *o, *a, *e (see correspondence 3c). Correspondences (3a) and (3b) apparently occur in the same environment, a fact that suggests that they contrast and therefore cannot be reduced to reflexes of a single segment; this pattern of contrast is only apparent, however, as I show now. As discussed in Carvalho and Rose (2018), PM had an (word-level) accentuation pattern characterized by left-aligned binary iambs, the exception being bi-syllabic words which instead show word-level accent in the first (word-initial) syllable, due to a general prohibition on accentuating the word-final syllable. With this pattern in mind, the following complementary distribution arises (the accented syllable is marked with an acute mark): for correspondence (3a) the PM reflexes of *ts are all in unaccented syllables *tsimápa ‘ashes’, *títsi ‘red’, *-mótsipa ‘eyelash’ (the latter is an inalienable noun, hence usually preceded by a possessive prefix, as in *nu-mótsipa ‘my eyelash’). For correspondence (3b), however, reflexes of *ts are always in an accented syllable: *iʧíni ‘jaguar’, *koʧíru ‘ant’. All three correspondences (3a), (3b) and (3c) can be resolved then as reflexes of a single segment *ts, which kept its affricate character in PM only in the context of a following *i and was palatalized to ʧ in a subset of this environment, that is, in accented syllables whose nucleus was *i10. The form *-ótse ‘grandmother’ is exceptional under this plausible account and is probably explainable by the action of some factor other than sound change. I will leave this to another paper devoted exclusively to the historical phonology of Mojeño11.

The reconstruction of *ts for correspondence (3c), an identity correspondence where all three languages agree in showing s, is perhaps the most controversial and the reasons for not reconstructing *s should be examined in greater detail.

Evidence for reconstructing *ts instead of *s comes, first, from the independently established fact that fricatives were debuccalized to h in Terena and that this language’s fricatives can be shown to come from affricates (Carvalho, 2017b, 2017c); hence, Terena s requires an affricate source. A second fact is that external evidence from other Arawakan languages supports the reconstruction of an affricate for the etymologies in (3a), (3b) and (3c). These are given below in (4) along with documentary evidence on early Terena showing the presence of affricates as well12:

(4) External evidence on the phonetic character of *ts
Woman: Paunaka esenu, PM *eseno and Terena sêno.
Palikur tino (Launey, 2003, p. 233), Campa (Asháninka and Ashéninka) tsinaneinane (Heitzman, 1973, p. 44), Proto-Arawakan *ʧɨna[ru] (Payne, 1991, p. 426).
cf. Early Terena <tseenŏ> ‘Weib’ (Schmidt, 1903, p. 570); <tséno> ‘Frau’ (Baldus, 1937, p. 539).
Mother-in-Law: Paunaka -muse, PM *-ímose, Terena -imóse.
Waurá -matɨ (Richards, 2015), Palikur -matru (< -matu-ru) (Launey, 2003, p. 234).
cf. Early Terena <imetse> ‘Schwiegermutter’ (Schmidt, 1903, p. 573).
Jaguar: Paunaka isini, PM *iʧini, Terena sîni.
Paresi ʧini (Rowan, 2001, p. 107), Proto-Arawakan *tsini PA (Payne, 1991, p. 409).
cf. Early Terena <tsiini> ‘Jaguar’ (Schmidt, 1903, p. 578).
Ashes: Paunaka sima-pa, PM *tsima-pa.
Campa (Asháninka, Ashéninka) tsitsi ‘fire’, tsimenkito/ʧimenkito ‘charcoal’ (Heitzman, 1973, p. 47-48), Paresi no-timi ‘my fire’ (Rowan, 2001, p. 99), Proto-Arawakan *tsɨma ‘Firewood’ (Payne, 1991, p. 403).

Moving on to correspondences (d), (e) and (f) of Table 6, repeated below in (5a-c) with the supporting cognate sets, note that the identity correspondence (5c), pointing to *t, was added due to its overlap with (5a).

(5) Correspondences and exemplar cognate sets for and *t
(5a) Paunaka ʧ: PM *ʧ : Ter t / __ *u
Head -ʧɨti : *-ʧuti : -tûti.
(5b) Paunaka ʧ: PM : Ter ʃ / __ *o,*e, *u
Know -iʧu : *-eʧo : -êʃo, Sun saʧe : *sáʧe : káʃe, Breast -ʧene: *-ʧene : -ʃêne,
Father-in-law -muʧɨku : *-ímoʧuko : -ímaʧuka.
(5c) Paunaka t: PM *t: Ter t /__*e, *i
Earth mute : *móte-hi : móte, Pain -kuti : *-koti : -kôti, Blood -iti : *-iti : -íti, Head -ʧɨti, *-ʧuti : -tûti,
Night juti : *joti : jóti , Red tisi : *titsi.

Correspondence (5b) is the main set for Paunaka ʧ which matches PM and Terena ʃ in the context of both back and front vowels. These can be understood as reflexes of , an interpretation consistent with the already mentioned independent evidence internal to Terena showing that the language changed its affricates to simple fricatives. Correspondence (5a) has to be recognized only for the unexpected correspondence t in Terena. If *t is assumed as the source, Paunaka -ʧɨti and PM *-ʧuti ‘head’ could be derived from the palatalization of *t triggered by *u, an outcome attested in some languages13. However, in all known cases, *t > ʧ triggered by a high back vowel is dependent on a high front vowel such as *i acting as a trigger as well (Bateman, 2011, p. 597) and, as seen in correspondence (5c), *i never triggers *t > ʧ in Paunaka or PM. I opt then for reconstructing for correspondence (5a), implying thus sporadic *ʧ > t in Terena.


Despite their restricted distribution, both correspondences in Table 7 and in (6) below contrast with each other and can be reconstructed as reflexes of distinct proto-segments *s and *h:

Table 7 Correspondence sets. 

Proto-segment Paunaka PM Terena
(a) *s s *h h
(b) *h h *h h

(6) Correspondences and exemplar cognate sets for *s and *h
(6a) Paunaka s: PM *h: Ter h / _ *i
Fingernail -sipu : *-hipoɲo : -hîpo, Tail -ke-isi : *-ihi-ki : -îhi,
Horn/Shoulder -siɨ : *-hiʔu.
(6b) Paunaka h: PM *h: Ter h / _ *i, *e, *u
Suck -uhiku : *-ohiko : -ohíko, Hair -hiju : -hijo-ʔo,
Moon kuhe : *kóhe : kohêe, Burn -ihɨe : *-íhu : -íhuwe, Grow -hɨku : *-huruko.

The two correspondences in (6a) and (6b) differ only in the segment found in Paunaka. Since both correspondences are attested preceding *i, two proto-segments are reconstructed and Paunaka turns out as more conservative than either PM or Terena, retaining the contrast between *s > s (6a) and *h > h (6b). Early written evidence on Terena is of pivotal importance here for two reasons. First, it shows that where Terena shows h in correspondence (6a) Early Terena had coronal fricatives instead, thus strengthening the case for a proto *s. Max Schmidt presents <šiipooti> ‘Nagel’ (Schmidt, 1903, p. 336), corresponding to modern hipô-ti ‘someone’s fingernail’; also, the name of a fish species given in Taunay (1875, p. 154), <Araraitti-issi>, literally ‘red tail’, is a compound whose first element matches modern hararáʔiti ‘red’, and whose second element shows a coronal fricative in the form for ‘tail’, <issi> (Modern Terena -îhi). Second, given the late date for the operation of the merger between *s and *h in Terena, we know that it was not a shared innovation of Terena and PM14.

Paunaka forms with s in correspondence (6a) above, which we analyze as reflexes of *s, are analyzed in Jolkesky (2016) as coming from *h instead. The author postulated ‘Pre-Paunaka’ forms showing *h, claiming that the attested modern Paunaka forms would derive from these by means of processes specific to this language. He claims explicitly that -siɨ ‘horn’ would come from Pre-Paunaka *-hiɨ by “[...] assimilation of a coronal feature [...]” (Jolkesky, 2016, p. 19), and a similar development is postulated for -isi ‘tail’, presumably from Pre-Paunaka *-ihi. Though such processes do exist, for instance, as synchronic allophonic realizations of an underlying glottal fricative, as in the well-known case of Japanese, e.g. hito ‘person’ [çito], the problem with applying this to a diachronic account of the correspondences in (6a) and (6b) is that it requires postulating a sporadic change of *h to s in the language, applying before *i in some cases (6a), but not in others (6b). Therefore, I reject this conclusion, recognizing instead a contrast *s - *h preserved in Paunaka.


Danielsen and Terhart (2014, p. 228) describe the rhotic r as a native phoneme in the Paunaka vocabulary, as opposed to the other liquid, the lateral l, whose occurrence is restricted to obvious loans15. Comparative evidence suggests, however, that r as well may be restricted to a loan stratum in the Paunaka lexicon, as *r was regularly lost in the inherited lexicon of the language, as seen by the correspondence sets in Table 8.

Table 8 Correspondence sets. 

Proto-segment Paunaka PM Terena Contexts
(a) *r *r elsewhere
(b) *r *r r ___*i

(7) Exemplar cognate sets for *r loss in Paunaka
(7a) Paunaka : PM *r: Terena
Pet peu : *péro : pêjo, Rib -himunepa : *-hirumonepa : -,
Grow -hɨku : *-huruko : -, Swallow -hikup- : *-hiriko- : -,
Body puɨ : *-poru : -.
(7b) Paunaka : PM *r : Terena r / __*i
Stone mai : *mári : marîpa.

Here, two correspondences (7a) and (7b) are recognized due to the varying reflexes in Terena, since Paunaka shows in all cases and PM retains *r >r16. The relevant conditioning factor seems to be the presence of a following *i in accounting for the preservation of *r in Terena. Though correspondence (7b) is attested in a single etymology this limitation stems probably from the relatively limited documentation available on Paunaka and from gaps resulting from lexical innovation, perhaps a result of borrowing. Comparison of Terena and PM cognates further supports the hypothesis that *i had a role in ‘protecting’ *r in Terena, though for these no Paunaka cognate exists: Terena jôu ‘fog’: PM *ijoru ‘fog’; Terena -ôo ‘to fly’: PM *-oro ‘to fly’; Terena -póe- ‘root’ : PM *-pore ‘root’, but Terena -kîri ‘nose’: PM *-siri; Terena -ámori ‘grandson’: PM *-amori17.

These correspondences suggest that forms having r in Paunaka are the result of borrowing, either from different, unrelated languages or from another dialect not subject *r > ∅. Aside from obvious loans from Spanish, such as bera ‘candle’, ajurau-ʧu ‘to help’ and arusu ‘rice’, many verbs, some cultural items and even some body-part terms have r and lack obvious cognates in the other languages: -bururuku ‘to boil’, -beriuk ‘to turn’, -kupuru ‘to get burned’, -arehik ‘to scratch, scrape’, -kurumehik ‘to pierce’, -kerahik ‘to break’, -marɨk ‘to cut’, -simirike ‘navel’, -kijuraki ‘brain’, kupisaɨrɨ ‘fox’, ubaramu ‘spider monkey’, upuri ‘snake’, takɨra ‘hen’, barereki ‘kettle, pot’. It will remain a task for future research to ascertain the source of this layer of loanwords in Paunaka.


The three overlapping correspondences in Table 9 provide evidence for the development of *k in Paunaka and its closest relatives.

Table 9 Correspondence sets. 

Proto-segment Paunaka PM Terena Contexts
(a) *k k *k k __ *o, *e
(b) *k k *s k ___*i
(c) *k s *s k ___ *a

(8) Exemplar cognate sets for *k
(8a) Paunaka k: PM *k: Terena k /__ *o, *e
Sleep -imu-ku : *-imo-ko : -imóko, Pain -kuti : *-koti : -kôti,
Ant kusijɨ : *koʧiru : kosîu, Moon kuhe : *kohe : kohêe, Soil apuke : *apokeʔe : pokéʔe.
(8b) Paunaka k : PM *s : Terena k /__*i
Wash -kipu : *-sipo-ko : *-kîpo, Turtle kipɨ : *sipu : -.
(8c) Paunaka s : PM *s : Terena k /__ *a /__ *a
Tapir samu : *samo : kámo, Sun saʧe : *saʧe : káʃe, Hear -samu : *-samo : -kâmo.

Correspondence (8a) above points unambiguously to *k. Correspondence (8b) shows the retention of *k as a velar stop in both Paunaka and Terena, while *k > s took place in PM where *i followed. Though the correspondences in (8) above are complementary and point therefore to the reconstruction of a single ancestral segment *k, correspondence (8c) raises two interesting issues: (1) an issue of phonetic motivation, given that *a is not an expected trigger for palatalizations, and (2) one of chronology, as the outcome is apparently shared between Paunaka and PM. These two independent issues are interrelated18.

Comparative evidence from more distantly related Arawakan languages shows that the low vowel *a in (8c) derives from *e or another front vowel, a more natural trigger of palatalizations19.

(9) External evidence on palatalization-triggering *a:
Tapir: Paunaka samu, PM *samo, Terena kámo.
Bahuana kema, Yucuna hema, Proto-Campa *kemari, Piapoco éma,
Mehinaku teme, Apurinã kema, Maipure <chièma>.
Hear: Paunaka -samu, PM *-samo, Terena -kâmo.
Palikur -timap, Baré -temuda, Bahuana -kimi-ta, Paresi -tsema,
Resígaro -heʔmɯ, Piapoco -émia-ka, Proto-Campa *-kema.

The data above suggests that *a was still a front vowel at some point in the relevant chronology, hence a natural trigger for the palatalization/coronalization *k > s. Later, this vowel merged with the low vowel *a, yielding the seemingly ‘odd’ phonetic conditioning pattern. Uncovering in detail how this came about and what where the reorganizations of the inherited vowel system that followed from this merger would take us beyond the scope of this paper, whose concern is solely the historical (segmental) phonology of Paunaka.

Finally, the question of whether the development *k > s was shared between Paunaka and PM – occurring, therefore, only once in a common period of development – or whether it took place independently in the two languages is also illuminated by comparisons with more distantly related Arawakan languages. Comparative data such as that in (9) suggests that the fronting/coronalization of *k preceding a front vowel took place many times independently in the Arawakan language family, an impression confirmed by the more general comparative study of Payne (1991), who claimed that the outcome is so general that some kind of allophonic split of *k and *kʰ in this context was a feature of Proto-Arawakan itself (Payne, 1991, p. 440). These facts point, in turn, to the high probability that Paunaka and PM may have innovated *k > s independently and that, therefore, this specific sound change bears little weight in supporting subgrouping arguments. The question of internal classification is the topic of the next section.


Table 10 presents three of the main phonological developments attested in the preceding sections for Paunaka. I compare Paunaka with Proto-Mojeño, Terena and Baure, the languages usually classified along with Paunaka in a close-knit subgroup variously labelled ‘Moxo/Moho’, ‘Bolívia-Paraná’ or ‘Southern Arawakan’ (Loukotka, 1968, p. 142; Kaufman, 1994, p. 59; Campbell, 1997, p. 181, 2012, p. 75; Danielsen, 2011, p. 517; Danielsen; Terhart, 2014, p. 224; Carvalho, 2017a). A cell in the table is marked with ‘yes’ anytime the development in question (broadly defined) is attested in the given language20

Table 10 Candidate shared innovations in Bolivia-Paraná historical phonology. 

*r > *u-fronting *k > s
Paunaka yes yes yes
Proto-Mojeño no no yes
Terena yes no no
Baure no yes yes

Paunaka shows the effects of the unconditioned loss of *r, while Terena retained *r > r in the context of a following *i. Though the two contexts are clearly different, it is not implausible to suppose that Paunaka may show the end-result of a generalization in the environment of a change that was less general, as seen in Terena. Nevertheless, in the absence of independent evidence that these two languages might be particularly closed related, that is, in the absence of other possible innovations shared by the two languages, it is hard to weigh in on the potential usefulness of this development for internal classification.

The change *u > ɨ underwent by Paunaka, if defined in purely phonetic terms as a fronting shift in the quality of the vowel, finds a parallel in a change that apparently also took place in Baure. Although Baure was not discussed in this paper, Table 11 below presents a very small set of etymologies that illustrates some of the phonological innovations of this language, so that these can be compared to those attested here for Paunaka21.

Table 11 Evidence on some Baure phonological innovations. 

Paunaka Terena Proto-Mojeño Baure PA
Water ɨne úne *úne in(e) *uni
Sky anɨ-mu wanúke *anu-mo ani *jenuh[kɨ]
Tapir samu kámo *samo som(o) *kema
Wash -kipu -kîpa *-sipo -sipa *kiba
Pain kuti -koti *-koti -koti *kaʧi[wi]
Eat -niku -nîko *-niko -nik(o) *nika
Go -junu -jôno *-jono -jon(o) *jani

Forms in Table 11 present evidence for a change *u > i in Baure (see ‘sky’ and ‘water’). Although this development is phonetically comparable to the *u > ɨ change in Paunaka, phonologically these two diachronic correspondences are very different: Baure *u > i apparently caused a merger of *u and *i, but the Paunaka shift *u > ɨ preserved a reconstructed contrast. The two changes are therefore clearly independent and have no weight at all in suggesting a closer relation between Paunaka and Baure22. Finally, as far as *k > s is concerned, Paunaka and Mojeño, as well as Baure, show this change in the context of a non-front vowel that can, however, be traced back to *e, as discussed in the preceding section. Baure and Mojeño agree, moreover, in showing the effects of *k > s preceding *i. As already pointed out these developments do not offer compelling evidence for subgrouping, as a coronalization/spirantization of *k in the context of a following front vowel took place many times independently in the Arawakan family.

It is legitimate, therefore, to look elsewhere, that is, in the lexicon and in the morphology, for candidate shared innovations, and these should be the focus of future work on the internal classification of Paunaka and its closest relatives. As mentioned before, there is in fact evidence from these domains suggesting that Paunaka and Mojeño, as well as Terena, to a lesser extent, may be particularly closely related. Jolkesky (2016, p. 27) presented several lexemes that seem to be shared between Mojeño and Paunaka, not Baure and Paikoneka, and advanced these as suggestive of a common period of development shared by Mojeño and Paunaka exclusively. As noted by Rose (2015a, p. 251), the third person non-specific prefix ti- of Mojeño has a plausible cognate in Paunaka ti- and these could be a shared innovation. Paunaka and Mojeño also seem to share a ‘relational noun’ used for the expression of possession with some nouns that cannot be directly marked by prefixes for possessor person/number, *-jeʔe in Proto-Mojeño (Rose, 2015b, p. 79; Carvalho; Rose, 2018) and -jai/-jae in Paunaka (Danielsen; Terhart, 2014, p. 236-237). While these observations raise the possibility that Mojeño may be the closest relative of Paunaka in the family, other studies have pointed out similarities found in Mojeño and in Terena. Carvalho (2017a, p. 82-83) notices a specific etymon, *-paho ‘mouth, opening’, that is shared between Terena and Proto-Mojeño but that seems to lack cognates elsewhere in the family, also adding additional observations on a shared pattern in the distribution of inherited roots and on one morphophonological effect of the suffixation of the first person plural suffix that seems to be unattested elsewhere. The overall conclusion of this section is that the phonological developments uncovered here for the history of Paunaka offer little in the way of support for specific hypotheses on the internal classification of this language. Nevertheless, the foregoing observations on certain lexical and morphological similarities involving Paunaka and Mojeño should be enough to make one more optimistic about finding potential shared innovations in these domains.


In this paper I have arrived at certain conclusions about the diachronic development of the segmental phonological system of Paunaka by applying the comparative method to a set of etymologies matching Paunaka forms to their cognates in Proto-Mojeño and in Terena. The relevant diachronic correspondences are given below in Table 12, while the set of etymologies is presented in the Appendix.

Table 12 Diachronic correspondences for Paunaka segments. 

Proto-segment Paunaka Proto-segment Paunaka
*p p *r
*t t *w ß
*ts s *j j
ʧ *a a
*k k, s *e e
*i i
*s s *o u
*h h *u ɨ

One of the most interesting findings is that the Paunaka vowel system featuring only a single back rounded vowel u is the reflex of a system that, just like those of Terena and Proto-Mojeño, was characterized by a contrast between two back rounded vowels *o and *u. Payne (1991, p. 476) claims that Chamicuro, Terena and Wayuunaiki are the only Arawakan languages with pervasive contrasts between u and o. Besides having no data at all on Paunaka at the time, Payne (1991) relied mostly on the Ignaciano dialect for his data on Mojeño, exactly the dialect of the language that has lost this opposition by merging PM *o and *a as a (Carvalho, 2017a; Carvalho; Rose, 2018)23. More complete cognate sets featuring Paunaka and PM, as well as Terena forms, allow one to demonstrate that such languages share a common ancestor that in fact had two contrastive back rounded vowels, thus showing that this contrast, of uncertain status in Arawakan historical phonology, can be reconstructed for an intermediate proto-language in the family.

Differently from PM and (Early) Terena (Carvalho, 2017b, 2017c), Paunaka lacks an alveolar affricate ts. Correspondences with these two other languages show that Paunaka merged *ts and *s in a context-free manner. Analysis of the etymologies addressed here reveals, however, that the contrast between these segments was somewhat limited and that *s had a very limited distribution restricted to the context of a following *i. Significantly, however, this is not an isolated property of the reconstructed system, as *t was also limited to the context of a following front vowel, either *e or *i. Finally, this agrees with some observations in Carvalho (2017b, 2017c) who, on the basis of documentary evidence on the contrast between fricatives and affricates in early stages of the history of Terena, indicates that while the contrast between ʧ and ʃ was clear and robust, this was not the case for the contrast between s and ts. Though segments with limited distributions and only marginal contrastive status are a cross-linguistic common feature, it is possible that further documentation of Paunaka may reveal a larger set of cognates that may, in turn, uncover a wider distributional range for these segments.

1I will refer to this family of clearly related languages by the label ‘Arawakan’, instead of the competing ‘Arawak’. See Michael and Granadillo (2014, p. 10) for this minor terminological quibble.

2<j> stands for a glottal fricative h and <ch> stands for the affricate ʧ in both modern and Old Baure.

3Note that the Paunaka back rounded vowel . has a lowered allophone [o] (Danielsen; Terhart, 2014, p. 229). Instances of . in available (phonologically transcribed) Paunaka data are restricted to unassimilated Spanish loanwords, such as comunidad (Danielsen; Terhart, 2014, p. 249) and amarillo (Danielsen; Terhart, 2014, p. 240).

5See e.g. Mihas (2015, p. 50) for the Alto Perené variety of Ashéninka.

6As all the compared languages show the variation [w] ~ [ß], the postulation of this phonetic shift is in a sense contingent on phonological analyses that may contain some element of arbitrariness. It is safer to postulate, however, the existence of a single contrastive unit and to suppose that the same allophony was found at the level of the proto-language.

7With regard to phonetic details, the two mergers are not entirely comparable since the Ignaciano merger would have involved an a rather close to o. Phonologically, however, both involve unconditioned mergers of non-front vowels. At any rate, the hypothesis of a merger between *o and *u as an explanation for the lack of contrastive back rounded vowels in Paunaka can be rejected.

8Note that this spurious correspondence showing PM *u: Paunaka u also includes Baure u, while Baure o appears in another correspondence. Quite strikingly, however, there is no contrast between u and o in Baure (Danielsen, 2007, p. 33).

9The fact that Terena was possibly once spoken in a region much closer to where the Paunaka live might suggest that the spirantization of *ts could have diffused from one language to the other, instead of being independent events. Documentary evidence on Terena discussed in Carvalho (2017b, 2017c) shows, however, that these developments took place very recently in Terena history, possibly around the turn of the 19th and 20th centuries, at a time when Terena speakers where already living in the eastern bank of the Paraguay river, within the territory of Brazil.

10Dependence of palatalization processes on prosodic structure is reported in Bateman (2011, p. 597) and references therein; see Giavazzi (2012) for the case of Italian.

11There are reasons to believe that the exceptional character of this form is due to phonosymbolism or analogy involving vocative or expressive parts of the kinship terminology. As discussed in Carvalho and Rose (2018), palatal ‘strengthening’ of certain vocative and ‘affective’ terms is attested in the development of the Ignaciano variety of Mojeño and in other branches of the Arawakan language family as well. Note also that in Terena there are pairs like se ‘grandmother’ but ôte/otête ‘granny’, ʃu ‘grandfather’, but ôtu ‘grandpa’ (also used as an affective address term for male elders in general).

12Also in this group one could place Terena -sêne: PM *-sene-ti ‘urine’, not discussed here because a cognate of this form seems not to exist in Paunaka. I take this to be an accidental gap, since cognates of these forms are widely attested elsewhere in the family and this is, to the best of our knowledge, a very stable PA etymon, reconstructed by Payne (1991, p. 424) as *ʧɨnɨ.

13See e.g. Stubbs (2000) on the Tepiman branch of Uto-Aztecan.

14See Carvalho (2017b, 2017c) for details on the Terena developments.

15e.g. asul ‘blue’ < Sp. azul ‘blue’; Danielsen and Terhart (2014, p. 255).

16The reader might ask whether the forms related to PM *pero ‘pet’ could be loans from Spanish perro. This is unlikely, as Payne (1991, p. 392) has shown that these are reflexes of a Proto-Arawakan etymon *pɨra ‘animal, domesticate’.

17Note that Paunaka -sɨsɨ ‘nose’ is not cognate with PM *-siri and Terena -kîri, both meaning ‘nose’ as well. In spite of the overall formal similarity, no regular correspondences support a match of Paunaka ɨ and PM *i / Terena i or of Paunaka s and PM *r / Terena r.

18Correspondence (8c) is interesting for yet another reason: in the absence of data from Terena, it is not possible to know whether Paunaka s : PM *s is a reflex of either *ts (as in 3c) or of *k (as in 8c). This is relevant, for instance, in the case of Paunaka samatɨ : PM *samatu ‘spider’ (Terena has the non-cognate form wáhaha ‘spider’).

19Sources on other Arawakan languages: Maipure (Zamponi, 2003), Yucuna (Schauer et al., 2005), Bahuana (Ramirez, 1992), Palikur (Launey, 2003), Paresi (Rowan, 2001), Resígaro (Allin, 1976), Apurinã (Facundes, 2000), Baré (Aikhenvald, 1995), Piapoco (Klumpp, 1995), Mehinaku (Corbera Mori, 2011), and Proto-Campa (Heitzman, 1973; Michael, 2011).

20One might wonder why Baure is excluded from much of the discussion in this article but has returned in the present section. The reason is that two separate issues or problems are at stake. Baure is less interesting than Mojeño or Terena for the task that constitutes the core of this paper – understanding the historical phonology of Paunaka – for reasons discussed above (its innovative character, lack of extensive lexical documentation etc.). For the specific issue addressed in the present section – the internal classification of Paunaka – a comparison with Baure is certainly relevant.

21The Proto-Arawakan (PA) forms are those of Payne (1991).

22It is true, as noted by Swintha Danielsen (personal communication), that a fronting change of the exact type as seen in Paunaka could have operated in Baure, as an intermediate step in the merger between *u and *i. Evaluating this hypothesis requires, however, a deeper investigation of the historical phonology of the Baure language, a topic outside the scope of the present article.

23A reviewer notes that a similar development took place in the Joaquiniano variety of Baure, as compared to Old Baure. This is surely an interesting topic, once more is known about the historical phonology of Baure and the diversification of its varieties.

24See Carvalho (2017b, 2017c) and references therein.

25The PM form means ‘leather, animal skin’. The classifier -mo denotes covering surfaces, so that the root *-poru arguably refers to what is covered by skin or leather.

26The change of w to j in Trinitario, preceding p, is a regular process (Gill, 1957, p. 15). Note that the terminus a quo for PM *-iwope is arguably *-iwo-pe ‘sole of the foot’, as shown by external comparanda such as Paunaka -ibu ‘foot’. The morpheme -pe is a classifier for flat or plank-like objects (Olza Zubiri et al., 2002, p. 275-277).

27PM has *-haka ‘mouth’, not clearly cognate.

28PM, Ignaciano and Trinitario forms mean ‘arm’. The classifier -ki qualifies cylindrical, rigid objects (Olza Zubiri et al., 2002, p. 288-302).

29The Terena form means ‘kidney’.

30Note that the Paunaka form in the etymology for ‘eye’ offers crucial evidence for *wu > u in PM. I thank Andrey Nikulin for pointing this out. See also that the etymology for ‘mosquito’ shows that a glottal stop is the reflex of *w in intervocalic position.

31PM *-huʔe means ‘belly; insides’, while Paunaka -hɨe-ki appears as a crystallized modifier meaning ‘insides’ or ‘inside part’, as in -ʧuka-hɨe-ki ‘inner ear, ear hole’, from -ʧuka ‘ear’ (cf. PM *-ʧoka).

32Paunaka esenu means ‘female animal’ (Spanish hembra).

33PM *(e)kuko is reconstructed based on Ignaciano -ékuka ‘uncle’ (Ott, W.; Ott, R., 1983, p. 633) and Old Mojeño <cuco>, <necuco> ‘my uncle’ (Marbán, 1701, p. 346). The Terena form is not obviously cognate because it requires the postulation of a sporadic loss of intervocalic *k. Crucial evidence exists, however, for the explanation of this as resulting from the analogical generalization of a morphophonemic alternation k ~ ∅. This fact belongs to the recent development of Terena and is not relevant for the present concerns.

34Paunaka and Terena forms mean ‘rain’, in PM and its daughter varieties the meaning is ‘cloud’. In PM, Ignaciano and Trinitario, -hi is a classifier denoting bulky, soft objects.

35See Paunaka mute-hi ‘earth, mud’, but mute-pa ‘earth, dust’.

36The suffix -pa is a classifier for powder-like objects (Olza Zubiri et al., 2002, p. 256-266).

37It is probable that an etymon *koe- ‘potato’ can be reconstructed and the Paunaka reflex shows the addition of the classifier -pi fossilized as part of the root. The cognate classifier in Mojeño also occurs many times with the nouns referring to plants (see api-pi kuhu ‘two yucas’, Olza Zubiri et al., 2002, p. 193). See also Paunaka for ‘Manioc’.

38Though a cognate of Paunaka ipiti is found, among Mojeño varieties, only in Old Mojeño, this constitutes likely a result of chance obsolescence/loss in the two extant varieties. An etymon close to *ipiti can be assumed for PM on the grounds of external evidence, from Paunaka, Terena and more distantly related branches such as the Campa languages.

39Terena -ohíko ‘to nurse (a child)’ is given in Ekdahl and Butler (1979). In my corpus the same root appears as meaning simply ‘to suck’.

40Paunaka final -e and Terena -we are not clearly cognate. Note for Terena, however, -íhuʃoa ‘to light fire for light’ (Ekdahl; Butler, 1969), where -ʃo is the thematic suffix and -a is an Object suffix, thus showing that -íhu- is synchronically segmentable.

41The Terena form shows the presence of a verbalizer/transitivizer prefix ko- and the effects of reduplication applied to a base *-(i)mo (see Rose, 2014 on reduplication in Mojeño).

42See the different thematic suffix in the languages: Terena and PM agree in showing reflexes of *-ʧo, while Paunaka has a reflex of *-ko (see cognate sets 102 and 103).


This Appendix presents 105 etymologies showing Paunaka forms and their cognates in Mojeño, Terena, or, in some cases, in only one of these languages. Of these, 99 are lexical items and 5 are grammatical/functional morphemes. A total of 101 Paunaka forms (96%) have a cognate in PM, while only 78 (74%) have a cognate in Terena. Some Terena forms appear between parentheses or followed by question marks whenever asserting their cognate status is less obvious or requires further elaboration in view of the regular correspondences discussed here. Data sources: Paunaka: Danielsen and Terhart (2014), Terhart (2014); Mojeño (Ignaciano): Ott, W. and Ott, R. (1983), Olza Zubiri et al. (2002); Mojeño (Trinitario): Gill (1957, 1993), Françoise Rose (personal communication), Carvalho and Rose (2018); Terena: author’s fieldwork data, Ekdahl and Butler (1969, 1979). I have tried to avoid as much as possible items that lie, by their meaning, outside basic vocabulary lists, unless they are particularly stable within the Arawakan family, show widespread cognates and are clearly reconstructible to much older stages, as is the case with many of the zoonyms such as Tapir, Monkey, Fish and Paca. For Mojeño, Old Mojeño forms from Marbán (1701) are at times introduced, between angled brackets and followed by OM (= Old Mojeño). Many of the Proto-Mojeño (PM) reconstructed forms below come from Carvalho and Rose (2018), but since the paper in question is not yet published I have included Ignaciano and Trinitario reflexes along with PM forms. Those PM etyma that are not in Carvalho and Rose (2018) are nevertheless entirely consistent with the correspondences identified by these authors and, given the minor differentiation of attested Mojeño varieties, rather obvious to establish. There are some morphological issues with these etymologies that will be skipped over in the present paper, but some commentary is offered in footnotes whenever necessary. As for transcription, all comparanda are entered in their phonological form and the symbols used, with few exceptions, have a standard IPA interpretation. Exceptions include, first, the rhotic, which in both Mojeño varieties and in Terena is usually a simple tap [ɾ] which is here represented as r. Also, in Terena, the acute and circumflex diacritics represent different accentual phonemes of the language24.

Cognate sets for Paunaka, Mojeño and Terena
Paunaka Proto-Mojeño Ignaciano Trinitario Terena
1 Body25 -puɨ *-poru-mo -paruma -prumo -
2 Head -ʧɨti *-ʧuti -ʧuti -ʧuti -tûti
3 Blood -iti *-iti -iti -iti -iti
4 Chin -mama *-mama -mama -mama -
5 Hand -ßuɨ *-woʔu -waʔu -woʔu -wôʔu
6 Foot26 -ißu *-iwo-pe -iwape -ijpe -
7 Mouth -nɨkɨ *-nuku -nuku -nuku -
8 Molar (tooth)27 -haka - - - -hâka
9 Tongue -pe-nene *-nene -nene -nene -nêne
10 Wing/Shoulder28 -pußu *-powo-ki -pawa-ki -pow-çi -pôwo
11 Horn/Shoulder -siɨ *-hiʔu -hiʔu -hiʔu -
12 Breast -ʧene *-ʧene -ʧene -ʧene -ʧêne
13 Back, lower29 -ʧupu-kekɨ *-keku -keku -çieku ʃúju-keku
14 bone -upe-hi *-ope-ra -apera -opera -ôpe
15 Rib -himunepa *-hirumonepa -hirumane -hiimonepa -
16 Fingernail -sipu *-hipoɲo -hipaɲa -hipɲo -hîpo
17 Eyelash -musipa *-motsi-pa -matsi -motsi-pa -
18 Hair -hiju *-hijo-ʔo -hija-ʔa -hijo-ʔo (-puhíʔo )
19 Eye30 -ßɨke *-uki-ʔa -uki-ʔa -uç-ʔa -ûke
20 Ear -ʧuka *-ʧoka -ʧaka -ʧoka -
21 Tail -ke-isi *-ihi-ki -ihi-ki -ih-çi -îhi
22 Meat/flesh -eʧe *-eʧe -eʧe -eʧe -
23 Belly (Inside)31 -hɨe-ki *-huʔe -huʔe -huʔe -
24 Wife -jenu *-jeno -jena -jeno -jêno
25 Husband -ima *-ima -ima -ima -îma
26 Brother -ati *-ati -ati -ati -âti
27 Son -ʧiʧa *-ʧiʧa -ʧiʧa -ʧiʧa -ʃeʔéʃa
28 Daughter -hine-pɨi - - - -ihíne
29 Woman32 esenu *eseno esena ʔseno sêno
30 Grandmother -use *-otse -atse -otse -ôse
31 Mother -enu *-eno -ena -eno -êno
32 Uncle33 kɨku *(e)kuko -ékuka <cuco, necuco> (OM) -eúko
33 Mother-in-law -muse *-ímose -ímase -imse -imóse
34 Father-in-law -muʧɨku *-ímoʧuko -ímaʧuka -imʧuko -imóʃuko
35 Water ɨne *úne une une úne
36 Sky anɨ-mo *anú-mo anuma anumo wanúke
37 Cloud/rain34 ɨku *úko-hi uka-hi uko-hi úko
38 Moon kuhe *kóhe kahe kohe kohêe
39 Sun saʧe *sáʧe saʧe saʧe káʃe
40 Night juti *jóti jati joti jóti
41 Stone mai *mári mari mari marîpa
42 Earth35 mute- *móte-hi mátehi mótehi móte
43 Soil apuke *apókeʔe apakeʔe ʔpóçʔe pokéʔe
44 Path -uʧene ~ -ʧene *-oʧene ~ -ʧene aʧene ʔʧene -oʃéne, ʃêne
45 Ashes36 sima-pa *tsima-pa tsima-pa tsma-pa -
46 Forest kimenu *simeno simena smeno -
47 Firewood (fire) jɨkɨ-ke *juku-ki juku-ki jkuçi juku
48 Garden asane-ti ‘field’ *esane-ti ésane-ti ésane isáne
49 Potato37 kuepi *koere kaere kaere koʔêe
50 Manioc kɨhɨ-pi *kuhu kuhu kuh-pa -
51 Pet -peu *-pero -pera -pero -pêjo
52 Fish himu *himo hima himo -
53 Snake keʧue *kiʧore kiʧare ççiore koéʃoe
54 Turtle kipɨ *sipu sipu sipu (Rose, personal communication) -
55 Tapir samu *sámo sama samo kámo
56 Capybara ɨʧɨ *uʧu uʧu uʧu -
57 Monkey iju *ijo ija ijo -
58 Paca jupu *jopo japa jopo -
59 Jaguar isini *iʧíni iʧíni ʔʧíni sîni
60 Duck upuhi *upóhi upahi ʔpohi pôhi
61 Wasp hane *háne hane hane háne
62 Bee, wax38 ipiti-umu <ypiti> (OM) - - pîti ~ -ípiti
63 Mosquito anißɨ *aniʔu aniʔu ʔɲiʔu nîu
64 Louse ine *-iɲe -iɲe -iɲe ɲâ-ti
65 Ant kusiɨ *koʧiru kaʧiru kʧiru kosîu
66 Spider samatɨ *samatu samatu smatu (Rose, personal communication) -
67 Bat ßite *wíte wite wite witête
68 Worm kane *kane (uʧepi) kane kâne
69 Be, stay -ußu *-owo -owo -owo -owo
70 Speak -keʧu - - - -kíʃo
71 Tell -kuetea *-koʔe -kaʔe -koʔe -kôʔe
72 Eat -niku *-niko -nika -niko -nîko
73 Suck39 -uhiku *-ohiko -áhika -ohko -ohíko
74 Laugh -ku *-eko-wo -ékawa -ekowo -éko-wo
75 Cry -ju *-íjo-ʔo -íja-ʔa -íjoʔo -íjo
76 Swallow -hikup-u *-hiriko -hirika -hiiko -huiri-ko (??)
77 Pain/hurt -kuti *-koti -kati -koti -kôti
78 Ripe -ju *-jo-ʔo -jaʔa -joʔo -
79 Red tisi *titsi- titsi- titsi -
80 Hit e-u *-e-ʔo -eʔa -eʔo -
81 Fall -ßenupu *-wenopo -wenapa -wenopo -
82 Grind -jɨßaika *-juwa-ko -juwaka -juwako júʃu ‘mortar’
83 Go -junu *-jono -jana -jono -jono
84 Grow -hɨku *-huruko -huruka -huuko -
85 Swell -amu *-amo -ama -amo -momoʔo-ʃo
86 Kill -kupa-ku *-kopa-ko -kapa-ka -kopa-ko -
87 Burn, catch fire40 -ihɨe *-íhu -ihu -íhu-ko ‘arder’ -íhuwe
88 Weed out (v.) -su *-iso-ʔo -ísaʔa -ísoʔo -íso
89 Dig -seku *-seko -seka -seko -
90 Know -iʧu *-eʧo -eʧa -eʧo -êʃo
91 Want, like -saʧ-u - - - -haʔáʃo
92 Fear -piku *-piko -pika -piko -pîko
93 See41 -imu *-imo-ʔo -ímaʔa -imʔo -komómo
94 Hear -samu *-samo -sama -samo -kâmo
95 Sleep -imu-ku- *-imo-ko -imaka -ímoko -imóko
96 Take -ße-u *-weʔo -weʔa -weʔo -wêo
97 Steal42 -umei-ku *-ome-ʧo -ameʧa -ómeʧo -omé-ʃo
98 Wash -kipu *-sipo-ko -sipa-ka -sip-ko -kîpo
99 Bathe -ku-ßu *-ko-wo -kawa -kowo -áhiko-wo
100 Defecate -suku *-soko -saka -soko -
101 Pron.Base -ti *-ti -ti -ti -ti
102 Absolute.Suffix -ti *-ti -ti -ti -ti
103 Thematic.Suffix1 -ku *-ko -ka -ko -k-o-
104 Thematic.Suffix2 -ʧu *-ʧo -ʧa -ʧo -ʃ-o-
105 Reciprocal -kuku *-koko -kaka -koko -koko


ADAM, Lucien; LECLERC, Charles (Ed.). Arte de la lengua de los Indios Baures de la Provincia de los Moxos: conforme al manuscrito original del P. Antonio Magio de la Compañía de Jesus. Paris: Maisonneuve y Libreros Editores, 1880. [ Links ]

AIKHENVALD, Alexandra Y. Bare. München: Lincom Europa, 1995. [ Links ]

AIKHENVALD, Alexandra Y. The Arawak language family. In: DIXON, R. M. W.; AIKHENVALD, A. Y. (Ed.). The Amazonian languages. Cambridge: Cambridge University Press, 1999. p. 65-102. [ Links ]

ALLIN, Trevor R. A grammar of Resígaro. 1976. 551 f. Thesis (Doctorate in Philosophy) – University of St. Andrews, Reino Unido, 1976. [ Links ]

BALDUS, Herbert. Tereno-Texte. Anthropos, Sankt Augustin, v. 32, n. 3-4, p. 528-544, Mai-Aug. 1937. [ Links ]

BATEMAN, Nicoleta. On the typology of palatalization. Language and Linguistics Compass, Hoboken, v. 5, n. 8, p. 588-602, Aug. 2011. DOI: [ Links ]

BENDOR-SAMUEL, John. An outline of the grammatical and phonological structure of Terêna. Brasília: Summer Institute of Linguistics (SIL), 1961. (Arquivo Linguístico, n. 90). [ Links ]

CAMPBELL, Lyle. Classification of the indigenous languages of South America. In: CAMPBELL, Lyle; GRONDONA, Verónica (Ed.). The indigenous languages of South America: a comprehensive guide. Berlin: De Gruyter Mouton, 2012. p. 59-166. (The World of Linguistics, 2). [ Links ]

CAMPBELL, Lyle. American Indian languages: the historical linguistics of Native America. Oxford: Oxford University Press, 1997. [ Links ]

CARDÚS, José. Las Misiones Franciscanas entre los infieles de Bolivia: descripcion del estado de ellas en 1883 y 1884, con una noticia sobre los caminos y tribus salvajes, una muestra de varias lenguas, curiosidades de historia natural, y un mapa para servir de ilustracion. Barcelona: Librería de la Inmaculada Concepcion, 1886. [ Links ]

CARVALHO, Fernando O. de. On Terena (Arawakan) -pâho ‘mouth’: etymology and implications for internal classification. Journal of Language Relationship, Moscow, v. 15, n. 2, p. 69-86, Quarterly 2017a. [ Links ]

CARVALHO, Fernando O. de. Fricative debuccalization and primary split in Terena (Arawak) historical phonology. International Journal of American Linguistics, Chicago, v. 83, n. 3, p. 509-537, July 2017b. DOI: [ Links ]

CARVALHO, Fernando O. de. Philological evidence for phonemic affricates and diachronic debuccalization in Early Terena (Arawak). Boletim do Museu Paraense Emílio Goeldi. Ciências Humanas, Belém, v. 12, n. 1, p. 157-171, jan.-abr. 2017c. DOI: [ Links ]

CARVALHO, Fernando O. de; ROSE, Françoise. Comparative reconstruction of Proto-Mojeño and the phonological diversification of Mojeño dialects. LIAMES, Campinas, v. 18, n. 1, p. 3-44, jan.-jun. 2018. [ Links ]

CORBERA MORI, Angel. Aspectos da morfofonologia e morfologia nominal da língua Mehinaku (Arawak). In: FRANCHETTO, Bruna (Ed.). Alto Xingu: uma sociedade multilingue. Rio de Janeiro: Museu do Índio/FUNAI, 2011. p. 193-216. [ Links ]

DANIELSEN, Swintha; TERHART, Lena. Paunaka. In: CREVELS, Mily; MUYSKEN, Pieter (Ed.). Lenguas de Bolivia: Oriente. La Paz: Plural Editores, 2014. v. 3, p. 221-258. [ Links ]

DANIELSEN, Swintha. Evaluating historical data (wordlists) in the case of Bolivian extinct languages. STUF - Language Typology and Universals, Berlin, v. 66, n. 3, p. 272-298, Oct. 2013. DOI: [ Links ]

DANIELSEN, Swintha. The personal paradigms in Baure and other Southern Arawakan languages. International Journal of American Linguistics, Chicago, v. 77, n. 4, p. 495-520, Oct. 2011. DOI: [ Links ]

DANIELSEN, Swintha. Baure: an Arawak language of Bolivia. Leiden: CNWS Publications, 2007. [ Links ]

EASTLACK, Charles L. Terena (Arawakan) pronouns. International Journal of American Linguistics, Chicago, v. 34, n. 1, p. 1-8, June 1968. [ Links ]

EKDAHL, Elizabeth Muriel; BUTLER, Nancy Evelyn. Aprenda Terêna. Dallas: Summer Institute of Linguistics, 1979. [ Links ]

EKDAHL, Elizabeth Muriel; BUTLER, Nancy. Terena dictionary. Dallas: Summer Institute of Linguistics, 1969. [ Links ]

EKDAHL, Muriel; GRIMES, Joseph E. Terena verb inflection. International Journal of American Linguistics, Chicago, v. 30, n. 3, p. 261-268, July 1964. [ Links ]

FACUNDES, Sidney. The language of the Apurinã people of Brazil (Maipure/Arawak). 2000. 705 f. Thesis (Doctorate in Philosophy) – University at Buffalo, Buffalo, 2000. [ Links ]

GIAVAZZI, Maria. Stress-conditioned palatalization in Italian. Manuscript. Massachusetts: MIT, 2012. [ Links ]

GILL, Wayne. Diccionario Trinitario-Castellano y Castellano-Trinitario. San Lorenzo de Mojos: Misión Evangélica Nuevas Tribus, 1993. [ Links ]

GILL, Wayne. Trinitario grammar. San Lorenzo de Mojos: Misión Nuevas Tribus, 1957. [ Links ]

HEITZMAN, Allene. Reflexes of some Proto-Campa consonants in modern Campan languages and dialects. 1973. 98 f. Thesis (Master in Lingustics) – University of Kansas, Lawrence, 1973. [ Links ]

JOLKESKY, Marcelo. Uma reconstrução do proto-mamoré-guaporé (família Arawak). LIAMES, Campinas, v. 16, n. 1, p. 7-37, jan.-jun. 2016. [ Links ]

KAUFMAN, Terrence. The native languages of South America. In: MOSELEY, C.; ASHER, R. E. (Ed.). Atlas of the World’s languages. London: Routledge, 1994. p. 59-93. [ Links ]

KLUMPP, Deloris Pharris. Vocabulario Piapoco-Español. Bogotá: Instituto Lingüístico de Verano, 1995. [ Links ]

LAUNEY, Michel. Awna Parikwaki: introduction à la langue Palikur de Guyane et de l’Amapá. Paris: IRD Editions, 2003. [ Links ]

LOUKOTKA, Cestmir. Classification of South American Indian languages. Los Angeles: University of California Press, 1968. [ Links ]

MARBÁN, Pedro. Arte de la lengua Moxa con su vocabulario y catecismo. Compuesto por el M. R. P. Pedro Marbán de la Compañia de Jesus, Superior, que fue, de las Missiones de Infieles, que tiene la Compañia de esta Provincia de el Peru em las dilatas Regiones de los Indios Moxox y Chiquitos. Lima: Imprenta Real de Joseph de Contreras, 1701. [ Links ]

MICHAEL, Lev; GRANADILLO, Tania. Introduction. In: MICHAEL, Lev; GRANADILLO, Tania (Ed.). Negation in Arawak languages. Leiden: Brill, 2014. p. 1-12. [ Links ]

MICHAEL, Lev. La reconstrucción y la clasificación interna de la Rama Kampa de la Família Arawak. In: CONGRESO DE IDIOMAS INDÍGENAS DE LATINOAMÉRICA, 5., Austin. Proceedings... Austin: University of Texas at Austin, 2011. [ Links ]

MIHAS, Elena. A grammar of Alto Perené (Arawak). Berlin: De Gruyter Mouton, 2015. (Mouton Grammar Library, v. 69). [ Links ]

OLZA ZUBIRI, Jesús; NUNI DE CHAPI, Conchita; TUBE, Juan. Gramática Moja Ignaciana: morfosintaxis. Caracas: Universidad Católica Andrés Bello, 2002. [ Links ]

OTT, Willis; OTT, Rebecca Burke de. Diccionario Ignaciano y Castellano con apuntes gramaticales. Cochabamba: Instituto Lingüístico de Verano (ILV), 1983. [ Links ]

PAYNE, David L. A classification of Maipuran (Arawakan) languages based on shared lexical retentions. In: DERBYSHIRE, Desmond C.; PULLUM, Geoffrey K. (Ed.). Handbook of Amazonian language. Berlin: Mouton de Gruyter, 1991. v. 3, p. 355-499. [ Links ]

RAMIREZ, Henri. Le Bahuana: une nouvelle langue de la famille Arawak. Paris: Amerindia, 1992. [ Links ]

RICHARDS, Joan. Waurá dictionary. In: KEY, Mary R.; COMRIE, Bernard (Ed.). the intercontinental dictionary series. Leipzig: Max Planck Institute for Evolutionary Anthropology, 2015. Available in: <>. Access in: 10 Dec. 2017. [ Links ]

ROSE, Françoise. Innovative complexity in the pronominal paradigm of Mojeño: a result of contact? In: GARDANI, Francesco; ARKADIEV, Peter; AMIRIDZE, Nino (Ed.). Borrowed morphology. Berlin: De Gruyter Mouton, 2015a. p. 241-267. (Language Contact and Bilinguism). [ Links ]

ROSE, Françoise. Mojeño Trinitario. In: CREVELS, Mily; MUYSKEN, Pieter (Ed.). Lenguas de Bolivia: Oriente. La Paz: Plural Editores, 2015b. v. 3, p. 59-97. [ Links ]

ROSE, Françoise. When vowel deletion blurs reduplication in Mojeño Trinitario. In: GÓMEZ, Gale Goodwin; VOORT, Hein van der (Ed.). Reduplication in South American languages. Leiden: Brill, 2014. p. 375-399. [ Links ]

ROSE, Françoise. Who is the third person? Fluid transitivity in Mojeño Trinitario. International Journal of American Linguistics, Chicago, v. 77, n. 4, p. 469-494, Oct. 2011. DOI: [ Links ]

ROWAN, Orlando (Comp.). Dicionário Paresi-Português. Cuiabá: Sociedade Internacional de Linguística, 2001. [ Links ]

SCHAUER, Stanley; SCHAUER, Junia; YUCUNA, Eladio; YUCUNA, Walter (Comp.). Diccionario bilingüe Yukuna-Español, Español-Yukuna. Bogotá: Editorial Buena Semilla, 2005. [ Links ]

SCHMIDT, Max. Guaná. Zeitschrift für Ethnologie, Berlin, v. 35, n. 2-3, p. 324-336, Jährlich 1903. [ Links ]

STUBBS, Brian D. More palatable reconstructions for Uto-Aztecan palatals. International Journal of American Linguistics, Chicago, v. 66, n. 1, p. 125-139, Jan. 2000. [ Links ]

TAUNAY, Alfredo d'Escragnolle. Vocabulario da língua Guaná ou Chané (Provincia do Matto Grosso). Revista Trimensal do Instituto Historico Geographico e Ethnographico do Brasil, Rio de Janeiro, tomo 38, parte segunda, p. 143-162, 1875. [ Links ]

TERHART, Lena. Paunaka field dictionary. Manuscript. 2014. Available in: <>. Access in: 25 Nov. 2017. [ Links ]

ZAMPONI, Raoul. Maipure. München: LINCOM Europa, 2003. [ Links ]

Received: November 14, 2017; Accepted: June 28, 2018

Autor para correspondência: Fernando O. de Carvalho. Universidade Federal do Amapá. Campus Marco Zero. Macapá, AP, Brasil. CEP 68902-280 ( ORCID:

Creative Commons License This is an Open Access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.