What are common origins of accusative case markers?

Question

What are the primary way(s) accusative case markers can develop?

The primary way of development of oblique case markers seems relatively straightforward, with attachment of adpositions (which can already be considered "case markers" under some definitions and which can in turn come from nouns or verbs) on the noun, then possible generalisation onto all elements of the NP as agreement. Ergative cases are a little less straightforward, but they very frequently have an oblique double function, hinting at origins such as reinterpretation of passive constructions or the generalisation of usage of e.g. instrumental or ablative case markers as markers of unexpected agency (e.g. via the occasional omission of agents in sentences like "[the man] opened the door with the key").

The emergence of accusative cases seem less straightforward to me though, and while reinterpretations of antipassives are a readily availible source that doesn't to me seem like a plausible origin for the majority of accusative markers, nor does it provide a way to go from no case marking to accusative (which seems like a thing that should be able to happen), and I'm not aware of any similar broad tendency of oblique double-function as seen in ergatives which could otherwise hint at origins. As such I particularly would like to know what oblique cases could reasonably be extended to accusative, and under what circumstances the original extension (from which further generalisation could be done) may occur under, and additionally if there are any reasons accusatives seem to have fewer double functions (even if this is just a result of me looking in the wrong places).

Answer 1

This is merely a marginal answer and I’m sure there’s a lot more data that might prove valuable, but in multiple Romance languages (at least Spanish and Romansh) the preposition a has developed into an accusative marker for particularly animate objects. This preposition derives from Latin ad “to(wards)” and is in both languages somewhat equivalent to English to or at in usage (in Romansh it is commonly used to mark the dative as well, as in French).

In general I would not be surprised to see allative or dative adpositions turn into accusative markers, there is quite a bit of overlap in the meanings (compare “I hit towards you” and “I hit you”).

Answer 2

Kuteva et al's The World Lexicon of Grammaticalization is the go-to academic source for this sort of thing.

---

The Second Edition lists four sources for the target PATIENT:

ALLATIVE: ways of marking motion towards can develop into marking for patients.

• Qiang: -ta developed from being an allative marker to being a patient marker
• Tibetan: -la used both allatively and as a patient marker
• Spanish: a, originally directional, then marked dative objects, and then direct objects that are animate nouns (especially individuated human nouns, poper nouns, and personal pronouns)
• Arabic: li- (classically) & *la-& (in some modern vernaculars) for certain direct objects (restrictions vary between dialects)
• Imonda: -m developed from a direction marker to an optional object marker but compulsory in [+HUMAN] object-subject relations
• Lezgian: -z "to" as a nominal suffix > -z as an experiencer object marker

They say there may be two pathways here, one via RECIPIENT, and another that goes to EXPERIENCER. They also note that there are grammaticalization pathways leading to ALLATIVE, so there will be multi-stage paths going via ALLATIVE.

Note that ALLATIVE should be understood in the broad sense here of ways of marking motion towards (including adpositions), and not solely to an allative case.

---

GIVE: verbs for giving can produce markers for patients.

• Qimen Hui: fã¹¹ developed from "give" to a differential object marker
• Southewestern Mandarin gěi is both a verb meaning "give" and an object marker
• The Xiang Dialects: use the verb pa³ as both a verb and object marker

Seems to be fairly common in Sinitic, but no examples listed outside Sino-Tibetan. They note that GIVE > RECIPIENT is pretty common, so this may be a multi-stage process.

---

RECIPIENT: marking for recipients (e.g. dative markers) can produce markers for patients.

• Maltese: human definite direct object generally receive the otherwise indirect object marker lil, indefinite human direct objects also sometimes receive it (but rarely), and inanimate direct objects never do regardless of definiteness.
• Arabic: dialectal la- "to, for" is a dative preposition used to mark certain direct objects.
• Hindi: the postposition ko "to" is used to mark animate direct objects.
• Dolakha-Newari: the dative case marker -ta developed into a patient marker.
• Old English: him was the dative 3rd person singular masculine, but in Modern English him is accusative/dative (likewise Old English hire and Modern English her)
• Spanish: a (see under ALLATIVE)
• Balti: la > -la
• Dhimal: e:ng > -e:ng
• Gurung: lai > -lai
• Tamang: ta > -ta
• Kupwar Kannada: dative postposition was extended to mark human direct objects due to contact with Urdu and Marathi

This typically involves contexts where there is a definite human referent to the patient, with it only later becoming generalised.

---

TAKE: verbs for taking can produce markers for patients.

• Medieval Chinese chi "take, hold" > chi instrumental preposition > chi patient marker
• Classical Chinese bă "take hold of" > Mandarin Chinese bǎ, an object marker
• Chinese jiang "take", "hold" > jiang a preverbal object (theme/undergoer) marker
• Formal Hakka, SOuthern Min, and Cantonese: jiāng "take, lead" > jiāng, an object marker
• Shanghainese Wu nɔ⁵³ "take, hold" > nɔ⁵³ an object marker
• Korean kaci- "take, have" > (ul)kac(i)ko, (ul)kacie which is an emphatic accusative
• Proto-Timor_Alor-Pantar *med "take" > Kamang me, Fatalkuku =m, -m a postposition encoding the displaced theme in a construction with a "give" verb.
• Kalam d "take" marks instrument or patient objects in specific contexts
• Engenni tọu "take" > object marker
• Vagala kpa "take" > object marker
• Ga kɛ̀ "take" > kɛ̀ an acccusative case marker
• Twi *de "take" > de an oject marker

There are also examples from pidgins and creoles, and it's well documented in Sinitic.

---

Note that the ALLATIVE & GIVE sources seem to also go via RECIPIENT, so really it seems that just the RECIPIENT and TAKE routes are independent.

In Addition, ALLATIVE & RECIPIENT can themselves develop from some other sources, listed below (in case you want more multi-stage options):

• ALLATIVE can develop from: ARRIVE, GO TO, SEE
• RECIPIENT can develop from: ALLATIVE, BENEFACTIVE, GIVE, A-POSSESSIVE (attributive possession "of", genitive case, associative or connective, as opposed to "belong"-possession or "have"-possession which are both predicative possession markers)