Tocantins Apiaká, Parirí and Yarumá as Members of the Pekodian Branch (Cariban)

This brief paper expands on the discussion of the evidence for including the sparsely attested Parirí and (Tocantins) Apiaká doculects as part of the Pekodian branch of the Cariban language family, within a sub-branch that includes the Arára-Ikpeng dialect cluster but excludes Bakairí. The present discussion goes beyond mere formal/semantic similarities in the comparanda and shows that both Parirí and Apiaká share a number of sporadic developments with Ikpeng and Arára, and these suggest an intermediate, shared ancestor exclusive to these languages. I also advance some original claims on the diachrony of these languages, such as the adoption of loans from non-Cariban languages, and an interesting semantic development in their innovative forms for ‘fire’. Based on this particular innovation, I conclude, tentatively, in favor of the inclusion of yet another doculect Yarumá in the same sub-branch.


Introduction 1
The labels Apiaká and Parirí appear in the ethnohistorical and anthropological literature on the indigenous peoples of South America associated with two Cariban-speaking groups inhabiting the region between the lower and middle courses of the Xingu and Tocantins rivers, to the south of the Amazon river, in northern Brazil. 2 Figure 1 below shows an inset of Nimuendajú's ethnolinguistic map of Brazil and adjacent regions, where the location of the relevant groups appear under the names 'Arára' and 'Parirí', both associated to the dates of the published reports identifying each of them:  (Nimuendajú 1987(Nimuendajú [1944).
1. I gratefully acknowledge an anonymous reviewer for his reading of this paper and for the relevant and fruitful commentary he/she offered on its contents. Needless to say, all remaining shortcomings this piece may have are of my entire responsability.
In the map above the Apiaká appear identified as the rightmost group identified as 'Arára', close to the left bank of the Tocantins river (the reasons for this will soon become clear to the reader). Since both ethnonyms, silently yet unproblematically understood as glottonyms, happen to be associated with language material (being doculects, sensu CYSOUW; GOOD, 2013), it seem fitting to examine these sources in an attempt to inform our understanding of both the internal history of the Apiaká and Parirí languages and, possibly, as an additional source on the external history of the Cariban language family. 3 As shown in section 2, the sources on Apiaká and Parirí have been examined in the past and a rather close proximity to Ikpeng, Arára and Bakairí, three independently attested Cariban languages that together make the Pekodian branch (MEIRA; FRANCHETTO, 2005), has been advanced. Published analyses of the data simply present a series of translational equivalents in a series of Cariban languages, with the goal of motivating an impression of greater proximity between Apiaká, Parirí and one or more of the Pekodian languages, as compared to other, less similar-looking Cariban languages (these were usually sampled from the lexical material published in Lucien Adam's well-known comparative study of the Cariban language family, published in 1893). In section 3, I will show that both Apiaká and Parirí have more than a set of similar lexical forms in common with the Arára-Ikpeng dialect cluster, 4 sharing with the latter a set of sporadic developments in specific lexical items. Since these are shared innovations exclusively characteristic of Arára-Ikpeng, this evidence establishes, once and for all, the Pekodian affiliation of both Apiaká and Parirí. In particular, they point to a rather close relation between Parirí, Apiaká, Arára and Ikpeng. Section 4 is dedicated to the conclusions of the paper.

The literature sofar
In his 1894 landmark publication Unter den Naturvölkern Zentral-Brasiliens, Karl von den Steinen compares some selected lexical material that he obtained among the Bakairí with the (then unpublished) Ehrenreich materials on the language of the Apiaká of the Tocantins river. Though his ultimate aim was to inform his hypotheses on the homeland and posterior migrations of Cariban speakers (see STEINEN, 1894, p. 395-404 for an interesting and pioneering discussion on the matter), he noticed pronounced similarities between the two languages. Steinen finds that these similarities are 'not as close as those observed between some Carib groups of the north of the Amazon and the Bakairi' (STEINEN, 1894, p. 400), 5 and concludes, based mainly on external evidence, that the Apiaká once lived close to the Bakairi, near the Paranatinga river and the upper Tapajós. In the nearly contemporary account of Ehrenreich (1895), Bakairi is also granted a special place in the comparison with Apiaká, if only because no data on the other Pekodian languages was available at the time. This changed with the publication of vocabulary data on both Parirí and Apiaká (NIMUENDAJÚ, 1914). After presenting his Parirí vocabulary, Nimuendajú (1914, pp. 624-625) adds very brief observations on the characteristic face painting and ornaments 6 of the Parirí, concluding that: "alles das bestätigt nur, was schon aus dem Vokabular hervorgeht: dass die Pariri eine Bande desselben Karibenstammes sind, der am Xingú als Arára, am Tocantins als Apiaká aufgetaucht ist" [All of this reinforces what was established by the vocabulary: That the Pariri are part of the same Carib group that appears in the Xingu as Arára and in the Tocantins as Apiaká] (NIMUENDAJÚ, 1914, p. 625). Nimuendajú (1914) explicitly proposes an identity between the Parirí, the Arára and the Tocantins Apiaká, all three being but regional individual communities (Bande) of the same group. Krause (1936) later brought to the picture the language of the Yarumá, another Cariban-speaking group that lived to the southeast of the Culuene river, one of the affluents of the Xingu. Krause (1936, p. 41) stressed the existence of a close relationship between Yarumá and (Tocantins) Apiaká. 7 This later, more complete view was enshrined in two important reference works, the chapter on the indigenous groups of the Xingu-Tocantins region in the Handbook of South American Indians (NIMUENDAJÚ, 1948), and the well-known Durbin (1977) classification of the Cariban language family. Durbin (1977, p. 31) added Ikpeng, then known as 'Txicão', to the same cluster. In the more ethnologically-oriented discussion of the Handbook, Nimuendajú (1948, pp.223-225) fleshes out in greater detail how the three regional subgroups of his 1914 quote given above emerged, introducing the Parirí and the Apiaká as two Arára subgroups that migrated eastward, the Apiaká eventually arriving at the left/west bank of the Tocantins river. Since the Apiaká are identified in Nimuendajú (1948, pp. 224) as 'eastern Arára', it is not surprising that they are identified simply as 'Arára' in his ethnolinguistic map (cf. Figure 1).
Later and more recent reference works do not diverge from this picture outlined in the pioneering contributions and, by their nature as reference works, do not examine the issue of the internal classification of these languages in any detail. In the Kaufman (1994Kaufman ( /2007 classification, Arara-Pirirí (that is, Arára and Parirí), Apiaká and Txicão (that is, Ikpeng) are all members of the same Arára group of his southern branch. Gildea (2012, p. 445) presents a 'Arara group' internal to his Pekodian branch, including Arára (Parirí), and Ikpeng, with Bakairí as a separate group within Pekodian. The same basic structure is repeated in the current Glottolog classification, which differs only by the inclusion of Yarumá in the group, by the lack of any mention to Parirí, and by the use of 'Xinguan' as a label for the Pekodian languages other than Bakairí.
6. The same comparison of tattooing patterns was employed by Eherenreich (1895) to ascertain the Cariban affiliation of the Apiaká of the Tocantins, bringing them tentatively close to the Arára, and to identify them as a group different from their Tupi-Guarani-speaking homonyms of the Tapajós river. 7. From the original: "Interessante und wichtig ist die enge Sprachverwandschaft mit den Apiaká des unteren Tocantins" (KRAUSE, 1936, p. 41). Yarumá will not be discussed here in any detail, given that this doculect, differently from Apiaká and Parirí, does not show the bulk of the relevant evidence for the shared innovations with the Ikpeng-Arára dialect cluster that are focused in the present paper. However, it will be briefly addressed again in section 4, since it seems to share with these languages one particular innovation.

Innovations shared by Apiaká, Parirí and Ikpeng-Arára.
As noted, the currently accepted classification of Apiaká as rather closely related to Arára-Ikpeng (MEIRA; FRANCHETTO, 2005, p. 130), and the inclusion of Parirí in this same group (GILDEA, 2012, p. 445), have been arrived at mainly by the impressionistic consideration of similarities in vocabulary material.
I will now discuss the data in table 1, showing that they constitute evidence that a set of sporadic innovations, most of which were identified by Meira & Franchetto (2005) as characteristic of Ikpeng are, first, also found in Arára and, second, also attested in Parirí and in Apiaká. See that these languages will be compared only with Bakairí in the table, as this is the closest relative of these languages, the other member of the Pekodian branch. Forms from other Cariban languages will be provided for comparison in the text below. Data from Apiaká come from Ehrenreich (1895), Parirí data from Nimuendajú (1914), Arara data from Alves (2010Alves ( , 2013 -m. 9 Syllable reduction is a well-known, but still poorly understood process that takes place in one form or another in every Cariban language (see GILDEA,1995, for a classic comparative discussion of the phenomenon). What matters for the moment is that Meira & Franchetto (2005), within the immediate purview of their reconstruction of Proto-Cariban phonology, quite correctly brush aside cases of syllable reduction as being matters of 'language-specific unpredictable idiosyncrasies' 9. The reader is referred to Meira & Franchetto (2005) for the full argumentation justifying the reconstruction of *w, and not *m, for the Proto-Cariban (PC) etymon for 'arrow'. This is, of course, vital, since if PC had *m instead here would be no Ikpeng innovation to begin with. The core reason for not reconstructing a nasal is that it would imply a series of ad hoc losses of nasal consonants in the other Cariban languages.  FRANCHETTO, 2005, p. 170). Be as that may, the fact is that the derivation of Ikpeng pɨrom 'arrow' calls for the postulation of a language-specific, idiosyncratic process of syllable reduction, followed by the operation of a still poorly understood development nasalizing *-w. There is, moreover, a strict relative chronology for these developments, as final vowel loss ('syllable reduction') is necessary to feed the operation of final *-w > m. The fact that Apiaká and Parirí, not to mention Arára, display cognates that call for the same analysis is a significant match, and one that applies to these languages alone among known Cariban languages.

Conclusions -and a note on Yarumá
This brief paper has brought to fore a series of innovations that single out a subgroup of Cariban languages -Ikpeng, Arára, Parirí and Apiaká do Tocantins -as a bona fide, if low-level, clade of this family. Though a modest contribution, it does succeed in establishing in terms of the methodologically more rigorous standard defined by the identification of sets of shared innovations what was once recognized only on the basis of impressionistic assessment of similarity. The developments discussed here, along with their distribution among the languages that make the Pekodian branch of the Cariban language family, are presented in table 2: Since each of the relevant doculects has its own particularities, some of which may derive from limitations in the existing records rather than from 'real' properties of the varieties they represent, they are best entered as separate labels in classifications, instead of having, say, Parirí as a variety of Arára, when it could, with equal justice, be described as a variety of Ikpeng. The structure I propose for this internal structure of the Pekodian branch is given in (1). The name 'Kampot' is here suggested, first, for its neutrality (as it does not adopt the label of one specific variety as a cover term for the sub-branch), and, second, for its convenience, since it is chosen after what is arguably one of the most interesting and distinguishing innovations setting this group of dialects/varieties from the rest of the family. Finally, note that Yarumá has also been included in (1) as part of the Kampot dialect cluster. The fragmentary attestation of this language, recorded by Wilhelm von den Steinen and Hermann Meyer, later published in Krause (1936), does not include forms for any of the meanings considered in table 1, thus being critically uninformative, with one exception. A form for 'fire' is noted: <kampón> (KRAUSE, 1936, p. 40). Since this form clearly fits within the same innovative set discussed in section 3, and since this innovation was singled out as the hallmark of this sub-branch, it is just natural that Yarumá should as well be recognized as part of this clade.