Table 1, which is organized by descending generation, summarizes Omagua and Kukama-Kukamiria forms denoting consanguines, together with the form reconstructed to POK. Consonants in parentheses are present in the Kukama dialect; boldface indicates forms with the same meaning but different etymologies; single horizontal lines separate generations. 8

POK distinguished parallel and cross relations in the first ascending generation. Cousin terms are left out of Table 1 because of incomplete data, although parallel and cross relations do not seem to have been distinguished in the ego’s generation, with sibling terms used for all cousins. Similarly, parallel and cross relations are not distinguished in the first descending generation, and nephew and niece terms are not dependent on the sex of ego’s siblings. At ego’s generation, as well as the first descending one, terms are distinguished based on ego’s sex. At the second descending generation a single grandchild term is employed, collapsing the sex of the referent relative to the second ascending generation.

Table 1. Omagua, Kukama-Kukamiria, and POK consanguines.

Table 1
Omagua, Kukama-Kukamiria, and POK consanguines

In Omagua there are three sibling terms, one for a same-sex sibling, imua, and distinct terms for opposite-sex siblings. In contrast, in Kukama-Kukamiria, there is a four-way distinction, with paja referring to a woman’s sister. In both languages, words for son and daughter differ minimally in the quality of the diphthong of the first syllable. Nephew and niece terms are derived from the corresponding direct descent terms (son and daughter) via the insertion of ɨ or i preceding the final vowel. The word for daughter is a compound based on the word for son. In Omagua, there is no distinction between daughter (female ego) and niece (female ego), although in Kukama-Kukamiria there is (see “Comparison with Tupinambá”).

The reconstruction of consanguineal terms to POK is relatively straightforward, and here I focus on those instances that require special explanation. For grandmother, the presence of a nasal vowel is unexpected in Omagua, as is the presence of an initial n in Kukama-Kukamiria where Omagua has no corresponding segment. TG languages typically have contrastive nasal vowels, but nasal vowels merged with oral ones in POK. 9

The terms for father and mother are probable borrowings, and were almost certainly borrowed into Omagua and Kukama-Kukamiria independently, given their divergence relative to the time of contact with Spanish speakers; for this reason I do not reconstruct them. The reconstruction of terms for parents’ siblings is similarly complicated. Omagua has terms for parents’ same-sex siblings that are based on the corresponding parent term in combination with the diminutive =kɨɾa; Kukama-Kukamiria has a similarly derived term only for maternal aunt, with pai referring to a paternal uncle. I reconstruct the asymmetry between parents’ same-sex sibling terms because a similar asymmetry exists in TG languages (with non-cognate forms), and because the Omagua term papakɨɾa is easily explained as an analogy based on mamakɨɾa, the reconstruction of irregularity is preferred. However, since the base on which mamakɨɾa is formed is itself a probable borrowing, I do not reconstruct an actual form for maternal aunt. In sum, the POK system thus appears to mirror a more widespread TG one in which the relevant forms have been replaced.

A similar asymmetry exists among sibling terms (such as Omagua imua and Kukama-Kukamiria paja), but in this case I reconstruct the Omagua system to POK, namely one in which *imua denotes a same-sex sibling, i.e., the brother of a man or the sister of a woman. Kukama-Kukamiria is consequently a borrowing from a still-unidentified language, since POK *imua is cognate to Tupinambá mu “[…] relative, ally, friend, nation, race […]” ( Barbosa, 1951, p. 103). A similarly broad range of meanings is found in other TG languages, thus a process of semantic narrowing appears to be at work, with the Tupinambá-like meaning at one extreme and the Kukama-Kukamiria one of brother (male ego) at the other. The Omagua meaning is intermediate, and as a result is likely to have been present in POK.

The reconstruction of terms in the first descending generation only requires additional explanation in the case of daughter and niece (female ego). In Omagua these two terms are identical, while in Kukama-Kukamiria they differ in the placement of stress on the second element of the compound, a reflex of *kunia ‘sister (ME),’ which is cognate to Tupinambá kuɲã ‘woman,’ which has the same meaning in most TG languages. 10 I suggest that the glide formation that *kunia underwent in Kukama-Kukamiria permitted the innovation of a term analogous to the other nephew and niece terms in these languages, namely the insertion of a vowel ɨ or i preceding the final vowel, 11 for example *memɨɾa ‘son (FE)’ vs. memɨɾɨa ‘nephew (FE).’ As a result, I reconstruct the polysemy found in the Omagua system to POK.

We now turn to a discussion of affinal terms in Omagua and Kukama-Kukamiria, which are summarized in Table 2. As before, single horizontal lines separate generations. 12

Table 2
Omagua, Kukama-Kukamiria, and POK affines

In ego’s generation there are distinctions based on ego’s sex, but in the ascending and descending generations there are no such distinctions. In the ascending generation, we see that *tutɨɾa and *jai tʃe are polysemous with parents’ opposite-sex siblings, suggesting a pattern of cross-cousin marriage. Recall, however, that terms for these in the first descending generation do not instantiate a parallel-cross distinction. Nor is such a distinction instantiated in affinal terms in the first descending generation: the terms for ‘son-in-law’ and ‘daughter-in-law’ show no similarities with ‘nephew’ and ‘niece,’ respectively, as might be expected given the polysemy in the first ascending generation. Finally, as in the consanguineal domain, opposite-sex in-law terms in ego’s generation are derived from the corresponding spousal terms via the insertion of ɨ.

The reconstruction of POK affinal terms is straightforward, with all pairs of forms descending from the same respective proto-forms. A small number of regular sound changes are attested (the lenition of * tʃ, the raising of *e, and glide formation). In addition, Kukama-Kukamiria underwent the insertion of the labiovelar glide w following u and preceding another vowel (cf. ‘daughter-in-law’).

The changes that affect consanguineal terms are summarized in Table 3. Underlined terms draw on TG lexical material in some way (see below); boldfaced terms are borrowings; and braces reflect the merging of two terms into one.

Table 3
Tupinambá and POK consanguines

We begin with a review of the changes in structure, all of which occurred in ego’s generation and in the two descending generations. The first concerns age grade distinctions among sibling terms. Tupinambá contained distinct terms for older and younger same-sex siblings, with opposite-sex siblings referred to with a single term regardless of age. In POK, these distinctions were leveled for egos of both sexes. The second concerns parallel-cross distinctions in the first descending generation. Tupinambá featured distinct terms for nephews and nieces depending on ego’s sex and whether the relation was parallel or cross. 13 Parallel nephews and nieces were referred to in the same way as children, with the addition of a compound variant term membɨkuɲã for a female ego’s parallel niece. Cross nephews and nieces were referred to with etymologically unrelated terms. Male egos distinguished the sex of the referent in such cases (cf. iɁɨɾ and jetipeɾ); female egos did not (cf. peη). All four pairs of terms were collapsed to four single terms, which were derived with the familiar insertion of ɨ or i. Furthermore, as a result of independent sound changes outside the scope of this article, daughter (female ego) came to be homophonous with niece (female ego), with subsequent innovation in Kukama-Kukamiria (see below). The third concerns an ego-based distinction in the second descending generation. In Tupinambá terms were distinguished based on ego’s sex and not referent’s sex, and these two terms were leveled in favor of the female ego form.

Now we turn to changes in the forms attested in this domain. In this vein, when I speak of the shift in meaning or replacement of a form in POK relative to Tupinambá, this is based on the widespread attestation of the relevant form with a particular meaning across many TG languages, which lies outside the scope of this article. The first change in form is the loss of a cognate to aɾɨj ‘grandmother’ and its replacement with *ãi and *nãi. Following sound changes attested elsewhere, these forms appear to be cognate with Tupinambá aɁi ‘mano’ (Portuguese) and naɁi ‘mana’ (Portuguese), respectively, which are familiar terms for brother and sister ( Barbosa, 1951, p. 24, 105). At this juncture it is important to highlight that nasal vowels generally merged with oral vowels in POK, as mentioned above. However, at least six forms with nasal vowels can be reconstructed for POK, as summarized in Table 4. These forms have nasal vowels in Omagua and oral vowels in Kukama-Kukamiria. Their cognates in Tupinambá exhibit uncommon phonological shapes in the language: a nasal vowel or diphthong preceded by a glottal consonant, or an oral diphthong preceded by a glottal stop.

Table 4
POK roots with nasal vowels

There is reason to believe that the relevant generalization might be that nasal vowels survived in POK if in the ancestral language they were the nucleus of a syllable that began with a glottal consonant (in all cases but one, the glottal stop): cognates of *tsɨi ‘tail’ in at least three TG languages exhibit the nasal diphthong ãj; cognates of *tsɨi ‘sweat’ do not exhibit a nasal diphthong, but the diphthong may have nasalized in the ancestor of POK by analogy with forms of similar phonological shape. There are exceedingly few forms in Tupinambá that contain syllables of the shape Ɂν(G), and even fewer that were inherited into POK; however, all forms with this syllable shape follow this pattern.

The above is intended to provide evidence for the cognacy of POK *ãi and *nãi ‘grandmother’ with Tupinambá aɁĩ ‘mano’ and naɁĩ ‘mana.’ At present, the specific trajectory of the semantic shifts involved is not clear: terms that are translated as informal sibling and friend vocatives came to refer to one’s grandmother. In that vein, I suggest that the vocative use of these terms in the ancestral language of Tupinambá and POK – namely, POKT – were likely different, allowing for a more natural-seeming semantic shift. The result is that two terms for grandmother were inherited into POK, with Omagua and Kukama-Kukamiria subsequently retaining different ones. Lastly, it is noteworthy that terms referring to women underwent considerable restructuring in POK, as summarized in Table 5, where the meanings of the Tupinambá forms are attested as such across the family. In that light, semantic shifts related to terms for grandmother are part of a broader set of shifts in this domain.

Table 5
Terms referring to women

The second change is the loss of the terms for father and mother. Terms for father and mother resembling papa and mama (see Table 1) are of course incredibly common, so common that it is likely that they were borrowed into Omagua and Kukama-Kukamiria independently. If they are of European origin, Spanish papá and mamá are likelier sources than Portuguese pai and mãe. Because of this likelihood, I do not reconstruct any such terms to P OK.

The term for paternal uncle, *pai, despite its resemblance to Portuguese, appearss to be cognate with Tupinambá paɁi, a vocative term for father ( Barbosa, 1951, p. 119). We have seen that Tupinambá uβ could refer to a father and to a paternal uncle, making it likely that paɁi was used as a vocative for the same individuals. In this case, this form came not only to be referential in POK, but also survives only in its reference to a paternal uncle. That is, the equation of a father with a father’s brother was split, perhaps under the influence of the borrowing for father. Related to this is the loss of the term for maternal aunt. Both Omagua and Kukama-Kukamiria have mamakɨɾa, but this form is based on an almost certain borrowing. Given that the enclitic *= kɨɾa was a diminutive in POK (see above), whatever the form is that preceded mamakɨɾa, it was likely based on a similarly morphologically complex term for mother, as in Tupinambá, such that the innovation of mamakɨɾa is a calque. Omagua subsequently replaced paternal uncle with papakɨɾa by analogy.

Moving on to terms in ego’s generation, we see that age grade-based sibling terms for male and female egos, a four-way distinction, were replaced by a single term, imua ‘same-sex sibling.’ This form is cognate to Tupinambá mũ, variously translated as relative, ally, friend, nation, and race ( Barbosa, 1951, p. 103). In the same domain, the term for a man’s sister was replaced due to the semantic shift of woman to sister. The only broader TG sibling term to survive as such is POK * kɨwɨɾa.

So far, what do these structural changes and changes in form tell us about the social situation in which POK developed? First, in the ascending generations, there seems to be a process whereby originally vocative terms came to replace referential terms. This might suggest that POK had cognates to Tupinambá father and mother vocative terms ( paɁi and aɁi ~ mãɁi, respectively) in referential function (i.e., possible * pai and * ai ~ * mai). In a situation in which POK developed in part due to widespread acquisition of a Tupinambá-like language by non-TG people, this would suggest that such speakers reinterpreted vocative terms as referential ones in overhearing the former. 14 In ego’s generation, two processes of semantic narrowing occurred (roughly, ‘relative’ > ‘same-sex sibling’ and ‘woman’ > ‘sister (ME)’). The former is perhaps explainable in terms of an increase in the expression of solidarity from the level of a family (or even extrafamilial group) to siblings.

Changes in form in the first descending generation are more straightforward. Terms for son (male ego) and daughter (male ego) were inherited directly. Terms for nephew (male ego) and niece (male ego) were derived via the insertion of ɨ and i, resulting in the loss of the parallel-cross distinction, as seen above. A slightly more complicated set of facts holds for daughter (female ego) and niece (female ego). For Tupinambá, Barbosa (1951) describes a single term for the former and two possible forms for the latter, with membɨkuրã being a compound of child and woman. I suggest that the same variation held for daughter, in which case membɨkuրã likely arose to distinguish between male and female children of female egos, as is possible with separate roots for male egos. If this is the case, the term was similarly inherited into POK in the same way as the other terms in the first descending generation. Subsequently, by analogy to sibling/nephew-niece pairs that alternated by way of the presence of ɨ or i, a distinction between daughter (female ego) and niece (female ego) was innovated in Kukama-Kukamiria ( Table 1). This was only possible in Kukama-Kukamiria, and not Omagua, because of the glide formation that has been discussed.

Generally speaking, there is more restructuring and more borrowing in the affinal domain than in the consanguineal domain, as summarized in Table 6. In terms of structure, in the first ascending generation ego-based distinctions are lost, making it similar to the lack of an ego-based distinction in this generation in the consanguineal domain. In ego’s generation, a six-way distinction in in-law terms for each of a male and female ego is reduced to a two-way distinction, but the ego-based distinctions are maintained. In the first descending generation, ego-based distinctions are lost in a way that exactly parallels the first ascending generation. This is unlike the consanguineal domain, where such distinctions remain. As a result, the consanguineal and affinal domains differ in the symmetry of ego-based distinctions across these three generations.

In terms of form, in this domain Tupinambá exhibits 14 compound forms involving two kin terms, out of a total of 20. Non-compound are aiʃo, men, emiɾeko, oβajaɾ, ukeɁi, and peɁum, which span all three generations. The compound terms are always transparent, e.g., menduβ, a combination of men ‘husband’ and uβ ‘father.’ Sometimes they attest to archaic forms, e.g., atuɁuβ, based on atɨ. In POK there are three borrowings, two internal derivations with ɨ, and semantic shifts that crosscut these two phenomena. Only *mena ‘husband’ and *miɾikua ‘wife’ were inherited with the same meaning.

Table 6
Tupinambá and POK affines

In the first ascending generation, terms for maternal uncle and paternal aunt ( Table 1) were extended to refer to parents-in-law. In ego’s generation, terms based on the spousal equivalents were derived via ɨ to refer to opposite-sex in-laws, that is, the brothers-in-law of women and the sisters-in-law of men. For same-sex in-laws, two processes are attested. For female egos, the TG form was simply retained (* uki); for male egos, the TG form was replaced by * ajuma, of unknown origin. And in the first descending generation, the resulting non-ego-based distinction is similarly expressed via two borrowings of unknown origin.

As in the consanguineal domain, we can ask what these linguistic changes tell us about possible social changes. The extension of uncle and aunt terms into the affinal domain suggests a pattern of cross-cousin marriage. 15 Elsewhere, it is significant that Tupinambá oβajaɾ referred to brothers-in-law as well as enemies – ‘ inimigo ( de nação), contrário’ (Portuguese) ( Barbosa, 1951). This polysemy almost certainly reflected a Tupinambá practice in which war captives were married to the capturing warriors’ sisters. The meaning of enemy was retained in POK * tsawajaɾa, a cognate. Similarly, Michael (2017) has suggested that the massive geographical and demographic expansion of speakers of pre-POK was tied to widespread captive-taking coupled with the abandonment of ritual cannibalism, which led to a large-scale influx and incorporation of speakers of non-TG languages into the POK population. When these speakers acquired pre-POK, their influences on its grammar and lexicon spread throughout the population. Regarding kin terms, and put somewhat differently, brothers-in-law of male egos would have ceased to be conceptualized as enemies, but rather incorporees into the society. Perhaps this is reflected in the borrowing of *ajuma. Nonetheless, it should be emphasized that borrowings in the first descending generation are not explainable in these terms, and that the different borrowings may naturally have had different motivations.

*Autor para correspondência: Zachary O’Hagan. University of California, Berkeley. Linguistics 1203 Dwinelle Hall. Berkeley, California 94720 USA (zohagan@berkeley.edu).