Two ga -morphemes in Rutul: Accidental similarity or a case of polygrammaticalization?

In a situation when two or more grammaticalization targets in one language are phonologically identical but functionally distinct, neither polygrammaticalization nor accidental syncretism can be ruled out, especially if we are dealing with a language without historical attestations. In the present paper, I present a detailed account of the coexistence of two homophonous grammatical markers in Rutul (< Lezgic < Nakh-Daghestanian), an underdescribed language of the Caucasus with a poorly known history. By taking into account the language-internal behavior of the markers, dialectal variation, similar developments in closely related languages and wider cross-linguistic parallels, I argue that the two ga-morphemes go back to constructions involving different lexical items and present the two different scenarios underlying their development. In particular, I suggest that the temporal ga and its variants originate from the noun ga(h) ‘time’, a Persian loan which also became the source of temporal markers in some other Lezgic languages. I also put forward a hypothesis that the indefiniteness marker ga comes from the verb ‘want’, a cross-linguistically common source for free-choice indefinite pronouns.