I have been trying for some time to figure out how to smoosh a bunch of my originally-unrelated languages into a larger family. They have a fair amount of noun and verb morphology that was suspiciously similar despite my not originally intending it. Here are some affixes I've reconstructed with their reflexes across 3 languages I'm trying to stitch together under this family: Proto-West-Celean (PWC), Mtsqrveli, and Apshur:

Proto-form PWC reflex Mtsqrveli reflex Apshur reflex Proto-meaning
*-VS *-oS (agent) -is (direct object), -os (nominalizer)? -os (nominalizer)? ???
*-Vm *-om (direct object) m- (verb transitive/telicizer)? /m/ in am (demonstrative)? definite???
*-l *-l-om (dative) -(Vl)i (genitive) -lda (benefactive) dative
* *-eH (feminine agent) -ɣe (benefactive) -a (oblique object), -lda (benefactive), -[w,h,d,z]a (allative) allative
*-r *-(e,o)r (transitive agent) -ar- (plural, earlier partitive) -Vr (ergative, genitive), -[wile,hile,di,zi]r (ablative) ablative
*-k *-(ey)k- (genitive) ? -kʰ (adjective ending) genitive
*-j *-ye- (verb stative aspect) -(Vl)i (genitive), -ia (definite agent), -is (definite direct object), iq'- (copula root) -di (pegative), -lda (benefactive), -ldi (instrumental), -da (superlative), -daj (superessive), -dir (superelative), -d- (verb subjunctive mood) stative
* *-ob-oS (comitative) ba (and), -[e,o]b (verbalizer) -wa (allative), -waj (adessive), -wiler (ablative), -Vw- (verb progressive marker) adessive ("next to")

You can see many of the reflexes are similar enough to reconstruct a proto-meaning. What's throwing me off is the cases where the PWC agent markers *-oS and *-eH correspond to object markers in the other languages. How to explain this discrepancy?

One idea I had was that maybe the parent *-VS and * morphemes simply had a meaning that could re-grammaticalize as either an agent or a patient - WLG 2019 implies the dative case (or, one step back, the genitive, ablative, or allative) could do this. But all of those already exist, and they all already map to different morphemes.

I also thought about how split-ergativity can arise in a nominative-accusative language, like in Hindi; verbs become rendered in the passive voice, ditching the semantic agent ("she(NOM) saw the man(ACC)" > "the man(NOM) was seen"), then the agent is reintroduced via an oblique case ( > "the man(NOM) was seen by her(INST)"), which gets reinterpreted as an active form (" > she(INST-ERG) saw the man(NOM)").

Apshur is purely erg/abs; PWC and Mtsqrveli are currently nom/acc, though I've been toying with making them split ergative instead (partially to explain how they could be sibling branches with a purely erg/abs language...). That + PWC seemingly inventing a new accusative case *-om without an equivalent in the other languages, suggests something like that ergativization process above might have happened, except... in reverse. But how does that process happen in reverse?

1 Answer 1


Straight up swapping the meaning of two case markers (let's call them like that for simplicity's sake) seems a pretty strange occurrence. There are however, theoretical ways this could happen:

  • You have some sort of "trojan horse" phenomenon with several words that have "inverted" arguments (cf. "I like X" vs. inverted Spanish "me gustas X" or German "Es gefällt mir"), and the inverted force somehow becomes generalized, causing what is essentially a reanalysis of these two cases.
  • Two of the main paradigm from two different cases became so similar as to generate confusion, and one of them is replaced with something else, which just happen to cause what looks like an inversion of the meanings of these affixes.

Though I know of no example offhand for affixes, I would be surprised if this second situation has not happened in some language somewhere! English borrowed they from Old Norse because the thirst person pronouns were getting almost all identical, she was also innovated for the same reason, but its sources are less clear (cross-linguistically, innovation or borrowing of pronouns is not a frequent phenomenon).

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.