The Evolution of Case in Two Otomanguean Languages of Mexico

Case marking systems often evolve diachronically from adpositions via a stage in which a former adposition now acts as a case-marking particle (Butt and King 2004, Rákosi and Laczkó 2011, inter alia). The correct analysis of such particles can be difficult to determine; do they continue to function as adpositions or have they now become instantiations of a syntactic Case part of speech? What are the intermediate stages of such a development like?

Two Otomanguean languages from Mexico -- Copala Triqui (CT) and San Juan Quiahije Chatino (SJQ) -- show case systems that shed light on two stages of the evolution of case systems from prepositions. Both CT and SJQ are head-initial languages with differential object marking for in-situ accusatives under similar conditions:

1a	San Juan Quiahije Chatino
	Ntqan42	Ma3rko24	qin4	Sa3na24/(qin4)	xneq2/(*qin4)	jyaq2.
	com:see	Marco	acc	Susana/(acc)	dog/(*acc)	soap
	'Marco saw Susana/the dog/the soap'

1b.	Copala Triqui
	Que-ne'en	Juán	(man)	Mariá/chuvee/yumi'
	com-see	Juan	(acc)	Maria/dog/soap
	'Juan saw Maria/the dog/soap'

Both the CT and SJQ accusative are sensitive to the definiteness and animacy of the object. CT shows optionality of the accusative in elicitation, but quantitatively accusative is extremely frequent with human and animal objects and very infrequent with inanimate objects. SJQ has grammaticized the same scale and requires accusative before human objects, prohibits it before inanimates, and makes it optional before animal objects.

Copala Triqui accusative as Case Of these two languages, the CT accusative is the one that shows the clearest signs of reanalysis as Case. Unlike other head-initial phrases in the languages (such as PP or NP), the CT accusative cannot be pied-piped when constituents are fronted for Contrastive or Interrogative Focus. Contrast 2(a,b) with (3a,b):¹

2a.	[PP Rihaan	chii]	achráá	Juán
	for	man	sing	Juan
	'Juan sings for the man.'
2a'.	¿[Me	síí	rihaan]	achráá	Juán?
		who	for	sing	Juan
	'Who does Juan sing for?
2b.	[NP Xnii	chii]	c-achráá.
	son	man	com-sing
	'The man's son sang.'
2b'.	¿[Me	síí	chii]	c-achráá?
		who	son	com-sing
	'Whose son sang?'
3a.	[CaseP	*Man	Mariá]	que-ne'en	Juán.
		acc	Maria	com-see	Juan
	'Juan saw Mary.'
3a'.	*¿[Me	síí	man]	que-ne'en	Juán?
		who	acc	com-see	Juan
	'Who did Juan see?
3b.	[NP Mariá]	que-ne'en	Juán.
	Maria	com-see	Juan
	'Juan saw Mary.'
3b'	¿[Me	síí]	que-ne'en	Juán?
		who	com-see	Juan
	'Who did Juan see?'

These facts fit much more readily in a grammar where man is a special part of speech, Case. Perhaps the easiest accommodation is via the PS-rules which license the initial focus phrase; they can specify that such phrases are limited to phrases of certain categories (NP, PP) but not others (CaseP, VP, CP).²

SJQ Monotransitives versus Ditransitives An initial examination of SJQ monotransitives like (1a) seems to show that the accusative marking in the language is purely prepositional, with the object in such an example showing a full range of phrasal behavior. However the account becomes more interesting when monotransitive SJQ clauses like (1a) are compared to ditransitives like (4) where qin4 has both accusative and dative functions. Here the evidence from the syntax shows that the dative qin4 is a Preposition, while the accusative qin4 has developed into Case:

4.	Yjwiq2	Sa3na24	qin4	Ti3la24	qin4	Tyu14+0.
	com:sell	Susana	acc	Tila	dat	Pedro
	'Susan sold (i.e. engaged) Tila to Pedro.'

Evidence for the Prep/Case distinction in SJQ As noted by Cruz et al. (2010), a true preposition like the qin4 marking recipients in SJQ allows two primary patterns in focus constructions; in both the preposition is stranded at the site of the gap. In variant (a), the focused element is alone in the filler slot, and in variant (b) a copy of the preposition accompanies the focused element.³

5.	Yjwiq2	Sa3na24	jyaq2	qin4	Tyu14+0
	com-sell	Susana	soap	to	Pedro
	'Susana sold soap to Pedro.'

6a.	Tyu14	yjwiq2	Sa3na24	jyaq2	qin1.
	Pedro	com-sell	Susana	soap	to
	'To Pedro, Susana sold soap.' (var A)
6b.	Qin4	Tyu14	yjwiq2	Sa3na24	jyaq2	qin1.
	to	Pedro	com-sell	Susana	soap	to
	'To Pedro, Susan sold soap.' (var B)

Cruz et al (2010) discover, but do not explain the fact that in contrast to this, the first object in a ditransitive has neither variant stranding pattern (7a,b). Its only possibility for contrastive fronting is as a bare NP without the accusative (7c):

7a.*	Ti3la24	yjwiq2	Sa3na24	qin4	qin4	Tyu14+0.
	Tila	com:sell	Susana	acc	dat	Pedro	acc
	[variant A]
7b.	*qin4	Ti3la24	yjwiq2	Sa3na24	qin4	qin4	Tyu14+0.
	Tila	com:sell	Susana	acc	dat	Pedro
	[variant B]
7c.	Ti3la24	yjwiq2	Sa3na24	qin4	Tyu14+0.
	Tila	com:sell	Susana	dat	Pedro
	'Tila, Susan sold to Pedro.'

The SJQ condition on accusative case on the first object in the ditransitive is thus identical to the restriction on the CT accusative CaseP. An overtly accusative marked CaseP does not participate in the Focus constructions of the language; only a bare NP does so. Note however that (7c) violates the otherwise exceptionless DOM rules by failing to mark a human object.

Ambiguity in monotransitives Returning to the monotransitive clauses in SJQ (1a), we may ask whether the qin4 in such clauses is also Case or whether it is a Preposition. For these clauses, the focus data are not distinct from the pattern seen with prepositions. The pattern where the object is fronted as a bare NP is ungrammatical:

8a.	Ti3la24	ntqan42	Mar3ko24	qin4.
	Tila	com:see	Marco	acc
	'Tila, Marco saw' (var A)
8b.	Qin4	Ti3la24	ntqan42	Mar3ko24	qin4.
	acc	Tila	com:see	Marco	acc
	'Tila, Marco saw.' (var B)
8c.	*Ti3la24	ntqan42	Mar3ko24.
	Tila	com:see	Marco

This might suggest that qin4 is always prepositional in such clauses. Yet the differential object data in (1a) point to an analysis of qin4 as Case, since the animacy of the object determines the presence of qin4 -- a relationship unlike that of a typical preposition to its object.

SJQ has already shown evidence of avoiding CaseP in focus. We argued earlier that in ditransitives, objects are realised as CaseP in situ as as NP in focus (4 vs 7c). For monotransitives, we propose that objects may be realised as CaseP in situ as as PP in focus, using an alternate analysis of qin4 as a preposition which assigns accusative. The failure of (8c) is due to the failure of a human object to receive accusative case. In the ditransitive, we hypothesize that the alternate analysis of qin4 as an accusative-assigning preposition is inhibited by the presence of a second qin4 in the same clause.

Thus the monotransitive qin4 is capable of two analyses in the grammar of SJQ. Its syntax has traits of both prepositional and case-like behavior that make it unlike any other word in the grammar. Both these Otomanguean languages show an intriguing restriction on the movement of CaseP that allows us to detect the full grammaticization of the Copala Triqui accusative and the grammaticization in progress of the San Juan Quiahije Chatino accusative.

End Notes

1. The a' and b' sentences also show some additional complications. Most interrogatives in CT are compounds of me, an interrogative element, plus a second element that shows the sortal type (animate, inanimate, place, time, etc.). Pied-piped interrogatives undergo a word order inversion which changes the normal head-initial order of these phrases. The second element of such inverted phrases undergo tone-lowering (marked by underscores).
2. Or perhaps a more insightful analysis would connect the limited movement possibilities of CaseP to other instances where Case needs to be licensed in specific configurations.
3. The correct analysis of the Variant (A) forms seems straightforward -- a stranded P and a focused object of P, but the best analysis of the variant (B) is far less clear.