Towards reconstructing the numeral classifier system of Proto-Tivoid

The Tivoid subgroup of Bantoid presents an evolving numeral classifier system with restricted lexical coverage, as attested for a number of various subgroups of the Benue Congo languages of Nigeria and Cameroon (Kießling 2018). Semantically, these classifiers categorise counted items for their shape and texture (e.g., oblong and rigid vs. flat vs. small and globular) as well as for their aggregation type (bundle vs. heap) and partition (half, piece) with an occasional conflation with the notion of counterexpectual scantiness. On the morphosyntactic and etymological level, they can be seen to develop from full-fledged generic nouns denoting concepts such as LEAF, SEED, FRUIT and HEAP used as head nouns in associative constructions. Eventual loss of nominal properties indexes an incipient functional split of the lexical source item and the newly emergent word class of numeral classifier. A comparison of numeral classifier systems in two Tivoid varieties, i.e. Tiv (Angitso 2020) and Ugare (Angitso & Kießling 2021), reveals both substantial overlap and variation. For example, cognate classifiers such as Tiv ítíné (5/6) and Ugare íʧín (5/6), both used for counting longish outgrowths from a base and applicable to items like plantains and hair, allow for a Proto-Tivoid reconstruction, whereas non-cognates such as Tiv ì-ké (9/6) ‘testicle’ vs. Ugare kù-kwà (9/10) ‘palm nut’, both used for counting items such as mangos and cashews, attest to the application of different cognitive models. Based on a comparison of the Tiv classifier system and its Ugare counterpart, the contribution explores the extent to which a numeral classifier system can be reconstructed for the Proto-Tivoid stage.