Comparing morphosyntactic patterns in Katukinan, Harakmbut and Arawan: Genealogical relations or language contact?

This paper investigates hypotheses about the genealogical relationship between the Katukinan (K), Harakmbut (H), and Arawan (A) language families put forward in the past two decades. Adelaar (2000) was the first to propose a genetic link between the Katukinan and Harakmbut families, adducing mainly lexical evidence (viz. 53 (near-)cognate lexical items); cognacy of grammatical items was scarce (only a handful of items). In his (2007) paper, Adelaar added data on Katawixí to the picture, also drawing on lexical evidence. Subsequently, Jolkesky (2016) put forward the addition of one more family to the genetic grouping, viz. the Arawan family, again on the basis of lexical correspondences.
However, morphosyntactic evidence for these hypotheses has been slim and unsystematic. This paper will further investigate this type of evidence in two ways. On the one hand, it takes a wide-scope approach, comparing grammatical descriptions of the three families. It will show that these families differ considerably in core aspects or areas of grammar. For instance, Katukina-Kanamari codes all TAM distinctions (except durative aspect) as clause-level particles (Queixalós Forthcoming), while Harakmbut codes them through suffixes attached to the finite verb (Van linden 2023). The marking of grammatical relations is also different. In Katukina-Kanamari, intransitive verbs go uninflected for person, as do transitive verbs if the subject participant is expressed by an external NP (Queixalós Forthcoming), as in (1). In Harakmbut, by contrast, all finite verb forms carry person prefixes, as in (2), with transitive verbs showing a hierarchical indexation system in which speech act participant O-arguments trigger relational person prefixes (Van linden 2022); compare oʔ-wek-me in (2) for a 3SG-acting-on-3SG configuration with mbeʔ-wek-ay-me-ne in (3) for a 3SG-acting-on-1SG configuration.

Katukina-Kanamari:
(1) pi:da-na= wu-tu niama tyohi.
jaguar-CASE= want-NEG then PalmSp.FruitBeverage
‘Then the jaguar did not want the palm sp.-fruit beverage.’ (Queixalós Forthcoming)

Harakmbut (Arakmbut/Amarakaeri variety):
(2) wa-mationka-eri oʔ-wek-me keme-ta
NMLZ-hunt-AN 3SG.IND-wound.with.arrow-REC.PST tapir-ACC
‘The hunter pierced the tapir.’ (Van linden 2023: 463)
(3) wa-mationka-eri mbeʔ-wek-ay-me-ne
NMLZ-hunt-AN 3SG>1/2SG-wound.with.arrow-AVRT-REC.PST-IND
‘The hunter almost pierced me.’ (Van linden 2023: 463)
Other areas in which these two families pattern differently include the marking of grammatical relations on dependents, adnominal possession, evidentiality, the coding of oblique participants, and noun incorporation. Such differences seem hard to reconcile with genetic relatedness.
On the other hand, we will compare inflectional and derivational formatives from the three language families, looking for grammatical cognacy much like the Katukinan-Arawan-Harakmbut Database (KAHD) (Gerardi et al. 2022, Gerardi et al. In preparation) seeks to identify lexical cognates. The languages included in this ‘grammatical’ database are Arakmbut/Amarakaeri (H), Deni (A), Jarawara (A), Kanamari (K), Kulina (A), Paumari (A) and Sorowaha (A). We also added Apurinã (Arawak) to the database, as it has been in contact with Katukinan and Arawan languages (Facundes 2000). These two ways of comparing morphosyntactic information, constructions and cognacy of morphemes, will allow us to shed more light on the question of whether these families are genetically related or if the similar lexical, morphological, and morphosyntactic patterns are due to horizontal transfer (borrowing of pattern and/or matter).