CC BY-NC-ND
11
Fernando O. de Carvalho
Federal University of Amapá (UNIFAP), Amapá State, Brazil.
Terena (Arawakan) -euko 'uncle' and -oko 'aunt': etymology and a kinship terminology puzzle1
This paper addresses the etymology of the nouns for 'uncle' and 'aunt' in Terena, an Arawakan language of Brazil. Analogy based on the model of a pattern attested in Old Mojeno explains a feature of the form -euko 'uncle' not accountable by regular sound change. The semantic side of the equations throws light on an aspect of the Terena kinship terminology that baffled anthropologists, supporting Oberg's conjecture that these kinship terms had their meanings extended to include parallel parental siblings. Finally, additional lexical reconstructions for Proto-Mojeno and an alternative analysis for an allomorphy pattern in Pa-resi are discussed as well.
Keywords: etymology; linguistic analogy; kinship terms; Arawakan languages.
1. Introduction
This brief paper offers an etymological analysis of two lexemes of the Terena (Arawakan) kinship terminology: -eúko 'uncle' and —ko 'aunt'.2 I single out these two specific items for detailed consideration for two reasons: within a properly linguistic domain of concern, these items, and, notably, the noun -eúko 'uncle', raise an etymological problem, as matching it with its obvious cognates in the closely related Mojeño would seemingly demand the acceptance of a sporadic, non-regular process of consonant loss. See (1) for the currently accepted classification of Terena and Mojeño within the same 'Bolivia-Parana' branch of the Arawakan family.3
1 Unless explicitly noted, all Terena data in this paper comes from the author's own fieldwork activities at the Cachoeirinha Reservation, Mato Grosso do Sul, Brazil. I am grateful to all my Terena consultants and friends for their patience and collaboration. I am also grateful to Ana Paula Brandao for discussion of data from Paresi and to an anonymous reviewer. All remaining flaws are my own.
2 The Terena phonemic inventory consists of the consonants p, t, k, m, n, p, s, J, r, l, w, j and the vowels a, e, i, o, u (see Bendor-Samuel 1961; Ekdahl, Butler 1979). A circumflex accent indicates greater length of the vowel where it occurs, in addition to a descending pitch curve (taki ['taiki] 'his/her arm'). The acute accent has no particularly salient pitch contour, and its lengthening effect is realized on the following consonant, not on the vowel above which it occurs (asurupi ['as:urupi] 'guts, intestines').
3 Some comments are in order: First, there is an emerging consensus that, within the Bolivia-Parana subgroup, Terena, Mojeno and Paunaka are closer to each other than any of these is to either Baure or Paikoneka (Jolkesky 2016; Carvalho 2017). Second, I have included early (17th century) documentations of Baure (Old Baure) and Mojeno (Old Mojeno) as dialects of the same language in order to avoid making the unnecessary and often indemonstrable assumption that these documents represent early stages of currently spoken varieties, as opposed to, say, extinct dialects whose speakers were either decimated by colonial action or simply shifted to some encroaching language (usually Spanish). The same applies to late 18th- and 19th-century sources on 'Guana', which, like Kinikinau and Layana, are co-dialects of the same language as Terena (see Carvalho 2016b for details).
Journal of Language Relationship • Вопросы языкового родства • 16/2 (2018) • Pp. 79-92 • © The authors, 2018
(1) Terena and its closest relatives Bolivia-Parana subgroup
Baure-Paikoneka branch
Baure (Baure, Old Baure, Joaquiniano) Paikoneka Achane branch Paunaka
Mojeño (Old Mojeño, Ignaciano, Trinitario, Javeriano, Loretano) Terena (Guaná, Chané, Kinikinau, Layana)
It can be shown, however, that instead of a sporadic process of consonant loss, simple analogical extension of a pattern observed elsewhere, namely in the kinship terminology of 17th century Old Mojeño, accounts for these developments. Finding a solution to this apparent difficulty constitutes a small yet real contribution to further understanding the historical development of Terena and its closest relatives.
The second reason, closely tied to the semantic side of the etymologies, stems from the puzzlement expressed by certain anthropologists that engaged in the study of Terena culture and social structure in the early decades of the 20th century. This was expressed by Oberg (1948: 287) who stated that:4
"(...) the Terena appear to have terms corresponding to uncle and aunt, for father's brother and mother's brother can be termed eungo or lulu, and father's sister and mother's sister can be termed ongo. A completely satisfactory explanation of these uncle and aunt terms cannot be made until more is known about the language and culture of the Terena."
The need for special explanation seems to stem from the following fact: Terena kinship terminology has transparent, descriptive terms for parallel parental siblings: poli nzâ?a 'father's brother' (lit. "my other father"; nzâ?a 'my father') and po?i ênô 'mother's sister' (lit. "my other mother"; ênô 'my mother'). Since -eúko 'uncle' and -ôko 'aunt' apply to both parallel and cross parental siblings, the superposition in Ego's parallel siblings, who can be denoted by either set of forms, is somewhat unexpected. Oberg's (1948: 287; 1949: 30) own suggestion is that -eúko 'uncle' and -ôko 'aunt' would be originally 'respect terms' for mother's brother and for father's sister, respectively, and that, at some point, these would have been extended in their use, just like the vocatives même 'mother (voc.)' and tâta 'father (voc.)' were extended to all older people of the parental generation. This scenario certainly has some interesting parallels, as in the case of the Iroquoian languages Huron and Wyandot, in which extension of a term for 'mother's brother' as referring to 'mother's sister's husband' was arguably facilitated by its previous use as a respect term used by younger men when addressing older men (Steckley 1993: 40-41).
I will argue that, although Oberg's (1948, 1949) intuition of a recent meaning extension in the reference of -eúko and -ôko is correct, these were not respect terms but were most likely the referential terminology used exclusively for cross parental siblings, later extended to include parallel parental siblings as well. The complex terms for parallel siblings are, in turn, recent formations that lack cognates even in the closely related Mojeño language.
4 The forms <eungo> and <ongo> given by Oberg (1948) are the first person singular possessive forms for -euko 'uncle' and -dko 'aunt', respectively. A first-person singular possessor is marked in Terena by a [nasal] feature that docks to the left edge of the word and spreads rightwards until it is blocked by an obstruent consonant. A short, transitional nasal consonant appears preceding the obstruent, which, in turn, becomes contextually voiced (see Carvalho 2017a). Hence: eungo 'my uncle', d"go 'my aunt'.
The paper is organized as follows: Section 2.1 deals with the formal issues raised by an attempt to relate Terena -euko 'uncle' and its Proto-Mojeno (PM) cognate *-ekuko 'uncle', arguing that analogical modification of a form *-ekuko 'uncle', reconstructed for a common ancestor of Terena and PM, accounts for the somewhat unexpected Terena form lacking a medial velar stop k. Section 2.2 briefly discusses evidence from a more distantly related language, Paresi, that is consistent with the reconstruction of *kuko 'uncle' (vocative) and *-ekuko 'uncle' (referential) proposed in section 2.1. Section 2.3. focuses on semantic issues. I argue that the reconstructed etyma *-ekuko and *-oko, until this point glossed simply as 'uncle' and 'aunt', respectively, had a more specialized meaning restricted to cross parental siblings only, that is: *-ekuko 'mother's brother' and *-oko 'father's sister'. I rely on the early attestation of these meanings in their Old Mojeno reflexes and on evidence from other Arawakan languages to support the postulation of these etyma. The complementary forms for parallel parental siblings attested in PM and in Terena are not cognate and were independently innovated in each of these languages.
2. Terena -euko 'uncle' and -oko 'aunt'
2.1. Formal issues
Comparison of the Terena noun -euko 'uncle' with its plausible cognate in Proto-Mojeno (henceforth PM), *-ekuko 'uncle', raises an etymological problem (see Carvalho & Rose 2018 for Proto-Mojeno phonology). Though the final syllable matches in accordance to regular (identity) sound correspondences, as well as the vowel e and the vowel u (see (2) below for some supporting cognate sets), a correspondence of PM *k to zero in Terena would require positing a sound change that lacks any motivation or independent support from regular developments.
(2) Identity regular correspondences for PM and Terena PM *e : Terena e
PM *-eno 'mother' : Terena -eno 'mother'; PM *-ope 'bone' : Terena -ope 'bone'; PM *-we-?o 'to take' : Terena -we(j)o 'to take'. PM *u : Terena u
PM *juku- 'fire(wood)' : Terena juku 'fire(wood)'; PM *-wo?u 'hand' : Terena -wo?u 'hand'; PM *une 'water' : Terena une 'water'. PM *k : Terena k
PM *-piko 'to fear' : Terena -piko 'to fear'; PM *koti 'pain' : Terena koti-we 'pain'; PM *apoke?e 'soil, earth' : Terena pokele 'earth, soil'. PM *o : Terena o
PM *-owo 'be, stay' : Terena -owo 'be, stay', PM *uko-hi 'cloud' : Terena uko 'rain'; PM *-jeno 'wife' : Terena -jeno 'wife'.
A solution to this problem is suggested by slightly broadening our perspective and including the cognate set for 'aunt' as well and, crucially, by looking at evidence from 17th century Old Mojeno (henceforth OM) in addition to the modern dialects of the language.5
5 Except for the Old Mojeno forms given between angled brackets, thus preserving the original orthography of Marban (1701), all forms from the Mojeno varieties were adapted from their source orthographies in agreement with IPA conventions. A source <y> is therefore adapted as j for the palatal glide, the allographs <c, qu> appear
Table 1. Forms for 'uncle' and 'aunt' in Terena and Mojeno6
Terena Ignaciano Trinitario Old Mojeno (OM)
'Uncle' -euko -ekuka (apiaru) -ekuko (ref.) kuko (voc.)
'Aunt' -oko -aka (apenru) -oko (ref.) koko (voc.)
Trinitario forms will not be relevant now, as sources on the language give forms for 'uncle' and 'aunt' that do not have cognates in Terena (I will come back to these Trinitario forms in section 2.3, where the semantic aspects of the relevant etymologies are discussed as well). In Ignaciano, according to Ott & Ott (1983: 632), -aka means 'sister of the grandmother', while -apenaru means 'sister of the mother'; -ekuka is given as a general term for 'uncle', in contrast to -apijaru, reserved for the father's brother (Ott & Ott 1983: 76, 633). There is no problem of either a formal or semantic nature in accepting that Ignaciano -aka is a cognate of OM -oko (and hence a reflex of PM *-oko) and of Terena -oko. On the formal side of the etymology, Carvalho (2017b) and Carvalho & Rose (2018) provide extensive evidence for the merger PM *a, *o > a in Ignaciano, which is also relevant in the match of Ignaciano -ekuka : OM -ekuko. The divergent semantics of Ignaciano -aka 'sister of the grandmother' will be briefly discussed in section 2.3, though a detailed treatment of this semantic mismatch will not be the focus of the present contribution.
The crucial dataset for addressing the etymological problem raised by Terena -euko is the OM data from Marban (1701), presented in table 1 in an adaptation of Marban's original orthography for the language (see Carvalho & Rose 2018 for the orthographic conventions of Marban). Marban (1701: 115-117) offers a discussion of kinship terms, carefully distinguishing vocative and referential forms (the latter described as 'possessive'). He notes, for instance, that <tata> 'my father' does not occur with the possessive prefixes; if these must be present, a separate form is used, as in <piya> 'your father' (with the second person singular possessor pi-), <maiya> 'his father' (with ma-, third person singular masculine possessor, for a male speaker)7 and <suiya> 'her father' (with su-, third person singular feminine possessor; see Marban 1701: 115). Likewise, for 'mother', where <meme> 'my mother' is distinguished from <peeno> 'your mother', <maeno> 'his mother' and so on.
Marban (1701: 115) presents <cuco> as meaning 'my uncle' and says that it admits the occurrence of person-marking prefixes, noting forms like<necuco> 'my uncle' and <pecuco> 'your (sg.) uncle'. However, these possessed forms call for the establishment of a distinct root, vowel-initial <-ecuco>, with the first person singular and second person singular prefixes, nu-and pi-, respectively, having their vowels lost to elision in internal sandhi with vowel-initial roots (see Carvalho & Rose 2018). This fact, in addition to the translation of <cuco> as 'my uncle',
here uniformly as k, and a glottal fricative is represented as h, not <j>. Ott & Ott (1983) also use <h> for a glottal stop, here represented as ?. Sources on the Mojeno varieties other than Marban (1701) usually employ an acute accentual mark signaling the syllable bearing word-level main stress. I have retained these when citing Mojeno forms.
6 In this section, mainly concerned with formal issues, I will refer to the meaning of the relevant cognate forms using the generic labels 'uncle' and 'aunt'. The distinction between cross and parallel uncle/aunt will be discussed in detail in section 2.3, where tables 4 and 5 offer a summary of the reconstructed etyma advanced here.
7 Third person singular pronouns in Mojeno index the sex of the speaker in addition to the gender of the referent. Thus, ma- is third person singular masculine for male speakers, while female speakers use pi- instead (see Rose 2015 for details).
despite the lack of any overt mark for a first person singular possessor, suggests the existence of two separate (albeit formally close) lexemes or terms: the vocative or address form <cuco> and a referential term whose root is <-ecuco>. Note also that there is no regular morphopho-nological process of vowel aphaeresis, in any documented variety of Mojeno, that could justify relating the free form <cuco> and the bound root <-ecuco> to the same underlying form, and I will therefore treat these as similar to the pairs <tata> (vocative) : <-iya> (referential) 'father' and <meme> (vocative) and <-eno> (referential) 'mother' noted above (see Carvalho & Rose 2018 for the reconstruction of *-ija 'father' and *-eno 'mother' for Proto-Mojeno). Finally, the same basic reasoning can be employed for recognizing two roots, one vocative and one referential, for 'aunt': the vocative <coco> is given by Marban (1701: 115) as meaning 'my aunt', whose gloss underscores its address function, and the possessive forms he cites, <nuoco> 'my aunt' and <pioco> 'your aunt'. These possessive forms, after the identification of the prefixes nu- and pi- call for the recognition of a root <-oco> 'aunt'. Again, as no regular process of elision of word-initial velar stops exists in Old Mojeno or in any Mojeno variety, the free form <coco> and the bound root <-oco> can be safely assigned to two independent terms or lexemes in OM. Finally, note that the distinct behavior of OM person-marking prefixes with the roots <-oco> and <-ecuco> follows from general properties of the morphophonology of the language: as noted in Carvalho & Rose (2018: 27), a prefix vowel in OM is regularly retained before a back (or non-front) vowel, as in <nuamori> 'my grandson' (Marban 1701: 289), but is lost preceding a front vowel, as in <nemotone> 'my work' and <nima> 'my husband' (Marban 1701: 502, 520), all with the first person singular prefix nu-.
Based on the OM evidence reviewed above and on cognates attested for Ignaciano (-ekuka 'uncle' and -aka 'aunt'), it is straightforward to reconstruct *-ekuko 'uncle' and *-oko 'aunt' for Proto-Mojeno. The two vocative forms found in OM, <cuco> 'uncle' and <coco> 'aunt', have no attested cognates in the other Mojeno varieties. Nevertheless, and this is the central insight offered here, reconstructing the pairs *-ekuko (referential) : *kuko (vocative) 'uncle' and *-oko (referential) : *koko (vocative) 'aunt' both for Proto-Mojeno, and for an earlier proto-language shared with Terena (possibly the 'Proto-Achane' level suggested in Carvalho 2017b), makes it possible to account for the unexpected k-less form -euko 'uncle' in Terena. Moreover, this account is also consistent with comparative evidence from other Arawakan languages. I will now deal with these two aspects of the diachronic account offered, reserving section 2.2 to the evidence from a more distantly related language.
The explanation I propose is based on an analogical modification of a form *-ekuko 'uncle', which is the expected, yet unattested, Terena match for Proto-Mojeno *-ekuko (see correspondences in 1 above).8 This analogy-based account is sketched in (2) below.
Table 2. Four-part (proportional) analogy underlying Terena *-ekuko > -euko
'Aunt' 'Uncle'
Vocative *koko *kuko
Referential *-oko *-ekuko > -euko
8 The crucial feature, for the present discussion, of the expected Terena cognate for PM *-ekuko 'uncle' is the presence of a medial velar stop, not present in the actually attested form -euko 'uncle'. When mentioning this expected yet unattested form I have retained the root-initial accentuation of the PM form. In fact, however, the nature of the comparative and diachronic relations between the PM and Terena prosodic systems remains unexplored and the identical position of the accentual marks in this case should not be seen as entailing a particular hypothesis on this matter or as having any implications whatsoever.
Table 2 above presents the set of referential and vocative forms for 'uncle' and 'aunt' that I reconstruct for some shared, intermediate proto-language whose set of daughter languages minimally include Terena and Proto-Mojeno. As seen above, this state of affairs is faithfully retained in 17th century Old Mojeno (OM). The specific analogical change that took place in Terena appears in the shaded cell in table 2. Terena -euko 'uncle' results from analogical modification based on the model implemented in the (arguably) semantically related forms for 'aunt': a vocative form with a kVkV shape matching a referential form with a VkV shape (*koko : *-oko). This proposal, which accounts for the Terena form which is otherwise surprising in view of attested sound correspondences with Proto-Mojeno, is entirely consistent with the traditional understanding of analogical changes as: "(...) a morphological transformation on the model of forms already existing in a language. When this occurs, purely phonetic developments in accordance with the sound laws are for the most part suppressed and obscured" (Szemerényi 1996: 27).9 Finally, note that the apparently reduplicated shape is a generalized property of vocative, respect and other address forms in Terena kinship and social relations terminology, including tâta 'father' and même 'mother' (cf. referential -ha?a and -êno, respectively) but also lûlu 'address term for male elders', ôtete 'address term for female elders' (cf. referential -ôte 'grandmother') and lêle 'address term for older male individuals of the same generation as Ego'. Thus, the reconstructed pair *kuko (vocative) : *-ékuko (referential), was not compliant with the general pattern having -CVCV vocative forms matching referential forms that lack this apparently reduplicated shape, and the simple deletion of the medial *k of the referential form brought this pair in line with this structural pattern, one that is transparently manifested in the semantically related pair *koko 'aunt' (vocative) : *oko 'aunt' (referential).
2.2. Evidence from Paresi-Haliti
1 have proposed that the set of forms in table 2, identical to those attested in OM, can be assumed for Proto-Mojeno and for some earlier proto-language ancestor also shared with Terena. The fact that assuming this set of forms allows one to explain the formally difficult Terena form -euko 'uncle' is arguably evidence for this. Nevertheless, as noted in section 1, Terena and Proto-Mojeno are very closely related and it would be good if evidence from other, more distant branches could offer additional support the reconstruction of the pattern in table
2 for a proto-language older than Proto-Mojeno itself. In Paresi (also Paresi-Haliti, Pareci), usually classified as either a 'Central Arawakan' (Payne 1991: 489) or 'Paresi-Xinguan' (Aik-henvald 1999: 67) language, one finds both koko, a vocative form for 'uncle' (as well as for 'father-in-law'; Ana Paula Brandao, p.c.)10 and the root -koke, the referential form for 'uncle' (Brandao 2014: 165). Besides, and perhaps of greater importance, -koke has a restricted vowel-initial allomorph -ekoke which occurs only with third person singular possessors. The restricted character of the allomorph -ekoke 'uncle' likely speaks for its inherited, primitive status, and hence for the existence, in Pre-Paresi, of the pair *-ekoke 'uncle' (referential) and *koko 'uncle' (vocative), a near-exact match to the situation attested in OM and projected back here as necessary to explain the odd Terena form -euko 'uncle'.
9 By 'morphological transformation' in Szemerenyi's quote one should understand 'formal modification', as opposed to looser definitions of analogical changes that include simple semantic shifts and other changes that lack formal repercussions under the same label of 'analogy'. See Hock (2003: 443-445) for discussion.
10 Polysemy involving 'uncle' and 'father-in-law' is certainly a reflection of a positive marriage rule with cross-cousins which is, in fact, found among the Paresi (see Florido 2008: 116-119). As seen in section 2.3 Paresi -ekoke ~ -koke denotes a cross-uncle, that is, 'mother's brother'. I keep the use of the simpler label 'uncle', however, since my main source on the language, Brandao (2014), also employs this generic gloss.
Before proceeding, additional commentary is necessary on the synchronic status of the Paresi -ekoke ~ -koke allomorphy. Brandao (2014: 128, 165), the primary source consulted here for this language, offers an analysis that differs slightly from mine: a single underlying form -koke 'uncle' is posited, with the vowel e of -ekoke, the restricted allomorph in my analysis, being assigned to the third person singular possessor marker, ene-, instead. Paresi has two alternants of the third person singular marker that are completely predictable on phonological grounds: e-preceding consonants (e.g. e-kahe 'his/her hand'), en- preceding vowels (e.g. en-eare 'his name'; see Brandao 2014: 164-165). For two items an additional allomorph ene- is postulated in Brandao's account: ene-koke 'his/her uncle' and ene-zona 'his/her ripe fruit'. Concerning the first of these forms, ene-koke 'his/her uncle', both my account and Brandao's require the specification of lexically-conditioned allomorphy: a distributionally very limited allomorph ene- of the third person singular marker in her account, a root allomorphy pattern -koke ~ -ekoke 'uncle' under my proposal. So, as they stand, both analyses invoke morphological idiosyncrasy, which seems in this case unavoidable and really demanded by the data. I think, however, that there are reasons for preferring the root-allomorphy analysis (-koke ~ -ekoke) over the prefix-allomorphy analysis (ene-, as well as en- ~ e-). Note, first, that comparative data from Proto-Mojeno and the likely shared ancestor of Terena and Mojeno speaks in favor of this solution, if not for a synchronic analysis of modern Paresi, at least for an early stage of the language, where Pre-Paresi *-ekoke 'uncle' would match PM *-ekuko. Second, and perhaps more critically, the other occurrence of the ene- allomorph in Brandao's (2014) account is likely amenable to an alternative analysis that eliminates the need to postulate a lexically-conditioned allomorph ene- for the third person singular prefix: ene-zona 'his ripe fruit' would be analyzable as en-ezona 'his ripe fruit' under the not far-fetched etymological equation with the independently attested verb -ezo- 'to fall' (Rowan 2001: 88), the semantic relation being established on the fact that ripe fruits are often identified as such once they fall off from trees. Still, a compromise or middle-ground solution is achievable if these options are understood as referring to different stages of the language: ene- could have developed in the modern language as a restricted al-lomorph after the sporadic absorption (reanalysis) of the root-initial e- of *-ekoke, whose existence is, after all, supported by comparative evidence. This will require delving into Paresi historical phonology and morphology, which lies outside the scope of the present contribution.
2.3. Semantics and the cross/parallel distinction among parental siblings
Concerning the semantic issues involving Terena -euko 'uncle' and -oko 'aunt', which relate to the ethnological problem brought up by Oberg (1948, 1949), the most important fact is that these terms are cognates of Old Mojeno terms for cross-uncle (mother's brother) and cross-aunt (father's sister), respectively. The distinction between parallel and cross terminology for Ego's parental generation is clear in the OM material of Marban (1701: 346), where the following forms are found: <Nuapenoru> 'Aunt, sister of my mother', <Coco>, <Nuoco> 'Aunt, sister of my father', <Nuapiyaru> 'Uncle, brother of my father' and <Cuco>, <Necuco> 'Uncle, brother of my mother'. Before discussing the semantics of the etyma for parental cross siblings ('father's sister' and 'mother's brother'), I will devote some space to the discussion of their complement, that is, the set of forms for parallel parental siblings, which are semantically unproblematic. Note that in this section I make use of the ordinary labels from kinship theory: FB= 'father's brother'; FZ= 'father's sister'; MB= 'mother's brother' and MZ= 'mother's sister'
Based on cognates in Ignaciano and Trinitario that deviate semantically to a small degree, Proto-Mojeno forms are reconstructed for parallel parental siblings as in table 3 below (these reconstructions constitute an addition to the existing corpus of reconstructed PM etyma appearing in Carvalho & Rose 2018).
Table 3. Proto-Mojeño terminology for parallel parental siblings
'father's brother' 'mother's sister'
Proto-Mojeño *-api-ija-ru *-api-eno-ru
Ignaciano -ápijaru -ápenaru
Trinitario -apiaru -apenru
Old Mojeño <Nuapiyarü> <Nuapenorü>
The Trinitario cognates are given only with non-specific labels for 'uncle', -apiaru 'tío' (Gill 1993: 2), and for 'aunt', -apenru 'tía' (Gill 1993: 2). For the Ignaciano variety, available sources confirm the restriction to parallel parental siblings, that is, 'mother's sister' and 'father's brother' or, respectively: -ápenaru 'tía (hermana de madre)' (Ott & Ott 1983: 74) and -ápijaru 'tío (hermano de padre)' (Ott & Ott 1983: 76). The matching to OM <Nuapenoru> 'Aunt, sister of my mother' (that is nu-apenoru), <Nuapiyaru> 'Uncle, brother of my father' (that is, nu-apijaru) is transparent and based on the semantics of the OM and Ignaciano cognates, I reconstruct *-api-ija-ru and *-api-eno-ru for 'father's brother' and 'mother's sister', respectively, in Proto-Mojeño.
These etyma include *-api-, the root for 'two', which occurs in combination not only with classifiers but in compounds with other lexemes, as in -ápiha 'surname', that is, -api-iha 'second name' (Ott & Ott 1983: 75), Trinitario api-miro 'hypocritical' (lit. "two faces"; Gill 1993: 2) and OM <apibé> 'two thorns, two hooks', <apicu> 'two rivers' and <apimo> 'two planting sites' (Marbán 1701: 381). From the evidence of all compared dialects it seems safe to reconstruct a synchronic morphophonological rule for PM that elided the final vowel of *-api- preceding either *-ija- 'father' or *-eno- 'mother'. The meaning of the two reconstructed etyma is thus purely descriptive, *-api-ija-ru 'father's brother' (lit. "second father") and *-api-eno-ru 'mother's sister' (lit. "second mother").
The fully transparent morphology and compositional semantics of these formations suggests a recent innovation. Evidence from the suffixal morphology of these etyma provides additional evidence for this later formation. The suffix -ru is a Nominalizer (Olza Zubiri et al. 2002: 626-641) which likely derives from earlier suffixes having both Nominalizing and Gender-marking functions, a property noted both in more general discussions of Arawakan morphology (see e.g. Payne 1987: 64) and in first-hand descriptions of these languages, as in Hanson (2010: 167-179) for Yine, Brandao (2014: 204-209) for Paresi and Pet (2011: 22-23) for Lokono. The function of these morphemes as Gender-markers was lost in Mojeño and in other members of the family (see e.g. Payne 1991: 377), but remnants of this use remain in a few unproductive corners of the morphology, as in Ignaciano máimaru 'widow, husbandless woman' (that is, ma-ima-ru; Ott & Ott 1983: 271; where -ima 'husband' and ma- is a Privative prefix) versus Ignaciano majenare 'widower, wifeless man' (that is, ma-jena-re; Ott & Ott 1983: 279; -jena 'wife') and --ha-ru 'name of a woman' vs. -íha-re 'name of a man' (Ott & Ott 1983: 219-220). Since in the case of the etyma *-api-ija-ru 'father's brother' and *-api-eno-ru 'mother's sister' the suffix *-ru is used with no regard for the Gender of the referent, their formation postdates the loss of the Gender-marking content of this morpheme, thus being a relatively late development.
In Terena, terms employed for parallel parental siblings, as anticipated in section 1, are also transparent formations: poli nza?a 'father's brother' (lit. "my other father"; nza?a 'my father') and po?i end 'mother's sister' (lit. "my other mother"; end 'my mother'). These are, clearly, not cognate with the PM etyma *-api-ija-ru 'father's brother' and *-api-eno-ru 'mother's
sister'. All in all, this is consistent with the idea that the terminology for parallel parental siblings was recently and independently innovated both in PM and in Terena. 11
Dealing now with the terminology for cross parental siblings, the central claim made here is that Terena -euko 'uncle' (that is, FB and MB), and -dko 'aunt' (that is, FZ and MZ) were subject to semantic broadening, since the etyma they derive from had their meanings restricted to 'cross uncle' (MB) and 'cross aunt' (FZ), respectively. Table 4 below shows the etyma *-oko 'father's sister' (FZ) and *-ekuko 'mother's brother' (MB) reconstructed for the common ancestor of Terena and PM, focusing on the semantic mismatches between their reflexes.
Table 4. Meanings of *-ekuko 'MB' and *-oko 'FZ' in Terena and two Mojeno dialects
Etyma Mojeño Terena
Old Mojeño Ignaciano
*-ekuko (MB) -ekuko 'mother's brother' (MB) -ékuka 'uncle' (MB, FB) -eúko 'uncle' (MB, FB)
*-oko (FZ) -oko 'father's sister' (FZ) -áka 'grandmother's sister' (FMZ, MMZ) -dko 'aunt' (MZ, FZ)
The semantic comparisons above lay out clearly a number of semantic diachronic correspondences. OM is conservative in retaining the semantic specialization to cross parental siblings of the reconstructed etyma. In Ignaciano, the sole modern dialect of the language where reflexes of these etyma are found, *-ekuko 'mother's brother' had its meaning broadened to include 'father's brother' as well. Ott & Ott (1983: 156) note that -ékuka is 'palabra general' ("generic word") for 'uncle', while -ápijaru is, as noted above, a transparent, special term for 'father's brother' (Ott & Ott 1983: 76). The etymon *-oko 'father's sister' changed to mean 'grandmother's sister' ('tía (hermana de la abuela)'; Ott & Ott 1983: 632).12 A reviewer questioned the inclusion of Ignaciano -aka 'grandmother's sister' in the same etymology as OM -oko 'father's sister' and Terena -dko 'aunt', possibly on the grounds that the diverging semantics precludes comparability. I think that this is not really a problem, but given this paper's focus on Terena, not on Mojeño, and the existence of certain unclear features in the meaning of the Ignaciano form (see footnote 12), I will comment only briefly on this. The postulated change from etymological *-oko 'father's sister' (FZ) to Ignaciano -aka 'grandmother's sister' (either FMZ or MMZ) consists, first and foremost, in ignoring or by-passing a generational difference in kin relations. This kind of semantic association is attested in many kinship systems and is, in fact, one of the defining properties of the Omaha/Crow terminological 'skewing' (Murdock 1949: 102; McConvell 2013: 154) where a single kinship term can refer to kin relations at Ego's generation (G0) and also at the parental generation (G+1),13 besides being also attested in diachronic se-
11 'Recently' here should be understood as meaning 'after PM and Terena split from their last shared common ancestor', which, as noted below, may be identified with the Proto-Achane level tentatively proposed in Carvalho (2017b).
12 The gloss provided by Ott & Ott (1983: 632) mentions no restriction to either FMZ or MMZ. Given the absence of any such specification, I have included a reference to either FMZ or MMZ as characteristic of the semantics of -aka.
13 In keeping with the traditional notation used in kinship theory, G0 denotes Ego's generation (e.g. his siblings and 'cousins', whether classified together with siblings or not), G+1 stands for the generation 'above' Ego, that is, the generation of his parents and their siblings, G-1 is the generation immediately 'below' G0, that is, that of Ego's children and the children of Ego's siblings, and so on.
mantic change, as in developments relating 'grandson' (G-2) and 'nephew' (G-1) in the Indo-European domain (see for example Mallory & Adams 1997: 239-240). Moreover, the fact that the sex of the connecting relative which was added as a function of the generational shift, in this case the grandmother (FM or MM), has changed (in the etymological meaning FZ the sole connecting relative is the male F) does not make the semantic association implausible, since the sex of the relative denoted by the term (the sister Z) is a more salient parameter in semantic change in kinship systems (see Hage 1999: 433) and this is kept constant in the change FZ > FMZ, MMZ. Given the typologically recurrent (even if 'marked'; see McConvell 2013: 154) status of semantic associations (or terminological classifications) that span that divide between generations in kin relations, and the fact that marking/salience relations proposed as constraints on diachronic change in kinship terms are not violated by this hypothesis, I think there are no obstacles of a semantic nature to the acceptance of Ignaciano -aka 'grandmother's sister' as a cognate of OM -oko 'father's sister' and Terena -dko 'aunt'.
For Terena, note that a single semantic broadening characterized by the loss of the cross/parallel distinction took place, that is: MB > (MB, FB) and FZ > (MZ, FZ). Terena -euko 'uncle' (MB, FB) and -oko 'aunt' (MZ, FZ) are the reflexes of etyma that were restricted in their reference to cross uncles (MB) and cross aunts (FZ). If the etymologizations in table 4 are indeed correct, I conclude that the semantic properties of the Terena terminology for parental siblings that intrigued Oberg (1948, 1949) stem from an extension in the meaning of forms traceable to *-oko 'father's sister' and *-ekuko 'mother's brother' and extension that derived a more 'classificatory' terminology referring to both cross and parallel parental siblings.
Given the variety of attested meanings in the terminology for parental siblings in OM, Ignaciano and Terena noted in table 4, it is perhaps necessary to discuss in greater detail the reasons for reconstructing the meanings *-ekuko 'mother's brother' (MB) and *-oko 'father's sister' (FZ), instead of, say, etyma with less specific meanings of the kind attested in Terena. In my view both the early attestation of the more specialized semantics denoting cross parental siblings in OM and comparative evidence from other Arawakan languages jointly furnish the required evidence for these semantic reconstructions and, therefore, for seeing Terena as definitely innovative. However, before discussing this in greater detail I will present in table 5 below a summary of the total set of forms for parental siblings reconstructed here for PM and for the shared ancestor of Terena and PM (identified here simply as 'etyma'). Note that cognates of Terena -euko and -oko in Proto-Mojeno are highlighted in bold, and so are the reconstructed etyma.14
Table 5. Terena and Proto-Mojeno terms for 'uncle' and 'aunt'
Meaning Terena Proto-Mojeño Etyma
FB -eúko *api-ija-ru
po?i -há?a
MB -eúko *-ékuko (ref.) *kuko (voc.) *-ekuko (ref.) *kuko (voc.)
FZ -ôko *-oko (ref.) *koko (voc.) *-oko (ref.) *koko (voc.)
MZ -ôko *api-eno-ru
po?i -êno
14 It is a well-known fact to linguists working on South American indigenous languages that a form comparable to kVkV and meaning 'uncle', where V stands usually for a back vowel u or o, is found throughout a number of unrelated languages and language groups, in particular in the Amazon. These include families like Arawá (Dixon 2004: 17), Panoan (Oliveira 2014: 417) and, of course, Arawakan.
As noted, both Terena and Mojeno have complex, derived terms for parallel parental siblings (MZ and FB); these, however, are not only not cognate, but are so transparent in formation that a recent origin, that is, one postdating their separation from the last shared common ancestor, can be safely ascribed to them. Note that the OM forms discussed in the present paper are conservative in that: (1) OM preserves both the form and meaning of the terms reconstructed for PM, including the distinct referential and vocative forms for MB and FZ, and (2) the specialized meanings of the PM forms for cross uncle (MB) and cross aunt (FZ), attested only in OM, are projected further back to the etyma reconstructed for the common ancestor of Terena and PM, tentatively identified with the Proto-Achane level suggested by Carvalho (2017b).
The fact that OM was attested some three hundred years earlier than Terena and Ignaciano is not in itself a strong reason for taking its semantics to be more conservative. It is rather easy, even trivial, to find cases in which languages/varieties of younger (more recent) attestation preserve more conservative or archaic structures — a phonological contrast, a specific morphological formation — which are nevertheless absent from languages/varieties of earlier (older) attestation (see Kümmel 2015 for a recent discussion). In fact, Carvalho & Rose (2018: 24) note that the OM variety described by Marban (1701) is less conservative than modern Ignaciano is in relation to the reconstructed accentual system of PM.
In the specific case of the semantics of kinship terms, however, I think that the early attestation of OM material makes it virtually certain that the meanings of these specific lexemes are more conservative than those attested much latter for Ignaciano and Terena. The same historic and cultural context that produced our existing documents on the Old Mojeno language, almost entirely restricted to the work of Marban (1701), was also characterized by a still significant preservation of pre-Columbian social institutions, attitudes, practices and behaviors, including marriage practices, residence patterns, religious beliefs and rituals of the Mojeno-speaking populations, all described to some extent in available sources (see Denevan 1966: 45-49; Saito 2015; Hirtzel 2016 and references therein). Given the well-known effects that "altered life conditions", to use Murdock's (1949: 199) apt phrase, have on the kinship systems (including kinship terminologies) of different peoples under the pressure of acculturation, notably as a result of changes on residence pattern, the influence of missionary activity and as a consequence of depopulation (see Eggan 1937: 39-40; Spoehr 1947; Murdock 1949: 199-202; Voget 1953; Balée 2014), it is much more probable than not that the modern Mojeno and Terena terminologies have been much more drastically affected by change, vis-à-vis their presumed etyma, than is the case with the kinship terminology attested for 17th OM. Though the relations between, on the one hand, kinship terminologies and, on the other hand, kinship systems and social institutions is a complex one, it would be surprising if the changes in demography and social structure to which native indigenous populations have been increasingly subject in the last centuries have failed to make the semantics of kinship terms in modern Mojeno and Terena communities less conservative and more innovative than the terminologies recorded from Mojeno speakers in the 17th century, when many of their pre-Columbian social institutions and belief systems were still preserved.
Finally, evidence from more distantly related Arawakan languages is also consistent with the hypothesis of an older kinship terminological system showing bifurcate merging in Ego's parental generation that is, one showing special terminology for cross parental siblings, or, in our case: *-ekuko for MB and *-oko for FZ. The Paresi term for 'uncle' discussed in the preceding section, -ekoke —koke, although often glossed simply as 'uncle' actually means 'mother's brother' (MB), as noted in ethnographic descriptions of the Paresi (see Bortoletto 1999: 58-59, fn. 27; Florido 2008: 117), the same being true for -nake 'aunt', or, more precisely, 'father's
sister' (FZ).15 The same ethnographies also add that the terms for 'father' (aba) and 'mother' (ama), in turn, classify both 'father' and his brother (F, FB) and 'mother' and her sister (M, MZ), respectively. Likewise, in Mehinaku, a language of the Xinguan branch (see e.g. Carvalho 2015, 2016a), papa classifies 'father' and 'father's brother', mama denotes both 'mother' and 'mother's sister' and specific terms refer to cross parental siblings: kuku 'mother's brother', aky 'father's sister' (see Florido 2008: 119-124 and references therein). Mehinaku -aky (FZ), Paresi nake < *na-ake (FZ) are plausibly cognates of the etymon *-oko (FZ) in table 5 above, and the same holds for Paresi -ekoke ~ -koke (MB), Mehinaku -kuku (MB) and the etymon *-ekoke (MB) proposed here. Unravelling the exact nature of these relations depends, however, on future comparative investigations aimed at working out the regular segmental correspondences matching, on the one hand, Terena and Mojeno and, on the other hand, Paresi and the Xin-guan languages.
3. Concluding remarks
This short contribution has provided an account to a specific formal difficulty arising from an attempt at relating the Terena noun -euko 'uncle' to its cognates in the rather closely related Mojeno language. Instead of invoking a sporadic and unmotivated process of consonant loss, an account grounded on the analogical imposition of a pattern attested in 17th century Old Mojeno offers a principled explanation for the occurrence of -euko instead of the predicted but unattested form *-ekuko. The etyma reconstructed for an intermediate ancestor shared by Terena and Proto-Mojeno, and some of its specific features, such as the co-existence of referential and vocative forms for 'uncle' differing only by the presence of anlaut vowel in the referential form, are, moreover, supported by evidence from the more distantly related Paresi.
As to the semantic changes to which the reconstructed etyma were subject, I have shown that *-ekuko 'mother's brother' (MB) and *-oko 'father's sister' (FZ) were subject to broadening changes in Terena, were their reflexes, -euko and -oko, respectively, also denote parallel kin relations. It was also shown that PM and Terena, which present cognate and specific forms for cross parental siblings have, nevertheless, independently innovated derived terms for parallel parental siblings. A reviewer suggests that this, combined with the evidence from other more distantly related Arawakan languages, such as Paresi and Waura, briefly discussed here, suffice as evidence for the hypothesis that Proto-Arawakan lacked specific terms for parallel parental siblings. I opt here, however, for a more cautious approach, given that Terena and PM are closely related languages and only very cursory comparison with a few other southern Arawakan languages has been presented here. Moreover, recent and pioneering comparative overviews of Arawakan kinship terminologies stress a high degree of variation and heterogeneity throughout the family (see Florido 2008: 160-161), a fact that makes even more risky any inference based on a few closely related languages. Perhaps Proto-Arawakan had complex, derived terms for parallel parental siblings of the kind seen in Terena and in PM, or, perhaps, a single underived form classified both the relevant parent and its sibling, as seen above in section 2.3 where Paresi aba 'F=FB' and ama 'M=MZ' were noted. Relating the reconstructed etyma presented here to their cognates in other Arawakan languages remains, therefore, a task for the future, once low-level (bottom-up) reconstruction has been successfully and extensively carried at the less inclusive level of closely related languages forming terminal branches.
15 Paresi -nake 'father's sister' likely includes a fossilized 1SG possessive prefix, *na-ake 'my aunt', and thus instantiates a vocative > referential shift.
References
Aikhenvald, Alexandra. 1999. The Arawakan language family. In: R. M. W. Dixon, Alexandra Aikhenvald (eds.) The Amazonian languages: 65-106. Cambridge University Press.
Balée, William. 2014. Charles Wagley on changes in Tupí-Guaraní kinship classifications. Boletim do Museu Paraense Emilio Goeldi 9 (3): 645-659.
Bendor-Samuel, John. 1961. An outline of the grammatical and phonological structure of Terena, Vols. 1 and 2. Arquivo Lingüístico 090. Brasilia: Summer Institute of Linguistics (SIL).
Bortoletto, Renata. 1999. Morfologia social Paresi: Uma etnografia das formas de sociabilidade de um grupo Aruak do Brasil central. MA Thesis, University of Campinas (UNICAMP), Brazil.
Brandäo, Ana Paula. 2014. A reference grammar of Paresi-Haliti. Doctoral Dissertation, University of Texas at Austin.
Cardoso de Oliveira, Roberto. 1976. Do indio ao bugre: O processo de assimilaçôo dos Terêna. Rio de Janeiro: Francisco Alves.
Carvalho, Fernando O. de. 2015. On the realization of nominal possession in Mehinaku: A diachronic account. International Journal of American Linguistics 81 (1): 119-132.
Carvalho, Fernando O. de. 2016a. Obscure cognates and lexical reconstruction: Notes on the diachrony of the Xin-guan Arawak languages. Boletim do Museu Paraense Emilio Goeldi, Ciências Humanas, 11 (1): 277-294.
Carvalho, Fernando O. de. 2016b. Terena, Chané, Guaná and Kinikinau are one and the same language: Setting the record straight on southern Arawakan linguistic diversity. LIAMES 16 (1): 39-57.
Carvalho, Fernando O. de. 2017a. Fricative debuccalization and primary split in Terena (Arawakan) historical phonology. International Journal of American Linguistics (july/2017 issue).
Carvalho, Fernando O. de. 2017b. On Terena (Arawakan) -pâho 'mouth': Etymology and implications for internal classification. Journal of Language Relationship 15 (2): 69-86.
Carvalho, Fernando O. de and Françoise Rose. 2018. Comparative reconstruction of proto-Mojeño and the phonological diversification of Mojeño dialects. LIAMES 18 (1): 3-44.
Denevan, William. 1966. The aboriginal cultural geography of the llanos de Mojos of Bolivia. Berkeley: University of California Press.
Dixon, R. M. W. 2004. Proto-Arawá phonology. Anthropological Linguistics 46 (1): 1-83.
Eggan, Fred. 1937. Historical Changes in the Choctaw Kinship System. American Anthropologist 39 (1): 34-52.
Ekdahl, Muriel, Nancy Butler. 1979. Aprenda Terêna. Vols. 1 and 2. Brasilia: Summer Institute of Linguistics (SIL).
Florido, Marcelo. 2008. As parentológicas Arawá e Arawak: Um estudo sobre parentesco e aliança. MA Thesis, University of Sao Paulo (USP), Brazil.
Gill, Wayne. 1993. Diccionario Trinitario-Castellano. San Lorenzo de Mojos: Misión Nuevas Tribus.
Hage, Per. 1999. Marking universals and the structure and evolution of kinship terminologies. The Journal of the Royal Anthropological Institute 5 (3): 423-441.
Hanson, Rebecca. 2010. A grammar of Yine. Doctoral Dissertation, La Trobe University.
Hirtzel, Vincent. 2016. Los espíritus-jaguares: cráneos-trofeos y chamanismo entre los Mojos (siglo XVII). Bulletin de l'Institut Français d'Études Andines 45 (1): 227-252.
Hock, Hans Henrich. 2003. Analogical change. In: Brian Joseph and Richard Janda (eds.). The handbook of historical linguistics: 441-460. Oxford / Berlin: Blackwell.
Jolkesky, Marcelo. 2016. Uma reconstruçao do proto-Mamoré-Guaporé (Familia Arawák). LIAMES 16 (1): 7-37.
Kümmel, Martin Joachim. 2015. Is ancient old and modern new? Fallacies of attestation and reconstruction (with special focus on Indo-Iranian). In: David Goldstein, Stephanie Jamison and Brent Vine (eds.). Proceedings of the 27th UCLA Indo-European Conference. Los Angeles, October 24th — 25th: 79-96. Bremen: Hempen.
Mallory, J. P., D. Q. Adams. 1997. Encyclopedia of Indo-European culture. Chicago: Fitzroy Dearborn.
McConvell, Patrick. 2013. Granny got cross: semantic change of kami 'mother's mother' to 'father's mother' in Pama-Nyungan. In: Robert Mailhammer (ed.) Lexical and structural etymology: 147-184. Berlin: Mouton De Gruyter.
Murdock, George P. 1949. Social structure. New York: The Macmillan Company.
Oberg, Kalervo. 1948. Terena social organization and law. American Anthropologist 50 (2): 283-291.
Oberg, Kalervo. 1949. The Terena and the Caduveo of Southern Mato Grosso, Brazil. Smithsonian Institution, Institute of Social Anthropology Publication N. 9. Washington: US Government Printing Office.
Oliveira, Sanderson. 2014. Contribuiçoes para a reconstruçâo do ProtoPáno. Doctoral Dissertation, Universidade de Brasília (UnB), Brazil.
Olza Zubiri, Jesus, Conchita Nuni de Chapi, Juan Tube. 2002. Gramática Moja Ignaciana (morfosintaxis). Universidad Católica del Táchira. San Cristóbal: Venezuela.
Payne, David. 1987. Some morphological elements of Maipuran Arawakan: Agreement affixes and the genitive construction. Language Sciences 9 (1): 57-75.
Payne, David. 1991. A classification of Maipuran (Arawakan) languages based on shared lexical retentions. In: Desmond Derbyshire and Geoffrey K. Pullum (eds.) Handbook of Amazonian languages. Vol. 3: 355-499. Berlin: Mouton de Gruyter.
Pet, Willem. 2011. A grammar sketch and lexicon of Arawak (Lokono Dian). SIL E-books 30. Summer Institute of Linguistics (SIL).
Rose, Françoise. 2015. Innovative complexity in the pronominal paradigm of Mojeño: a result of contact? In: Francesco Gardani, Peter Arkadiev and Nino Amiridze (eds.) Borrowed Morphology: 241-267. Berlin: Mouton de Gruyter.
Rowan, Orlando. 2001. Dicionário Paresi-Português. Cuiabá: Sociedade Internacional de Lingüística.
Saito, Akira. 2015. Guerra y evangelización en las misiones jesuíticas de Moxos. Boletín Americanista, Año LXV, 1(70): 35-56.
Spoehr, Alexander. 1947. Changing kinship systems. Anthropological Series Volume 33 Number 4. Chicago: Field Museum of Natural History.
Steckley, John. 1993. Huron kinship terminology. Ontario Archaeology 55: 35-59.
Szemerényi, Oswald. 1996. Introduction to Indo-European linguistics. Oxford: Oxford University Press.
Voget, Fred. 1953. Kinship changes at Caughnawaga. American Anthropologist 55 (3): 385-394.
Фернанду де Карвалью. Морфемы -eúko 'дядя' и -ôko 'тетя' в языке терена (аравакская семья): этимологическая разгадка одной аномалии в системе терминов родства
В статье предлагается оригинальная этимологизация имен существительных дядя' и 'тетя' в бразильском языке терена, относящемся к аравакской семье. В частности, некоторые особенности развития формы -eúko дядя', не имеющие объяснения в рамках обычной теории звуковых законов, предлагается считать аналогическим преобразованием по той же модели, которая засвидетельствована в старом мохеньо. Семантическое сопоставление вероятных когнатов позволяет пролить свет на странности в терминологии родства у терена, давно озадачивавшие антропологов, и подтвердить старое предположение Оберга о том, что сфера применимости соответствующих терминов родства в какой-то момент распространилась и на параллельных сиблингов. По ходу дела в статье также представлен ряд дополнительных лексических реконструкций для языка прамохеньо и альтернативный вариант анализа системы распределения алломорфов в языке пареси.
Ключевые слова: этимология, языковая аналогия, термины родства, аравакские языки.
