Necesito algo más atípico... no sé muy bien qué!

La cosa... flexiona por sujeto/objeto... pero la marcación de casos va un poco a su bola ^^

Source - ergative, ablative, genitive

Current - subject of motion, DO of giving, perlative/translative

Recipient - accusative, IO of giving, allative 

Locative - space-temporal location and S, possessive

//Instrumental - Instrument, some causative subjects??

Partitive - Quantified predicates, group of superlatives, maybe second of comparatives?

//Absolutive - Intransitive subject

To qualify as core (that is, to reflect onto verbal semantic inflection), predicate must be animate. Gender is syntactic, though. Reflects the animacy and number of the primary object/intransitive subject argument. 

No inanimate as transitive subject, ever. Well, their status is unclear.

So:   I carved some figurines -> figurine(i).ptv some.rec 1-carve(i)-agt

      I hunted some deer -> deer(a,pl).ptv some.rec 1-hunt(a,pl)-dir,v

      You hurted me -> 1-hurt(a,sg)-inv,nv-2

      A rock fell -> rock(i,pl).loc fall(i,pl)

Affectedness... it seems to be the key.

This would be incorporated, usually... except for emphasis.

Now... I think I should distinguish where a body part acts as an instrument... mmm.

I'm going to merge locative and instrumental. So five cases, but unlike many systems I know. Bueno, no... en la raíz están claramente diferenciados, y en los aplicativos será distinto también. Así pues, seis casos. No están demasiado ligados al tema sintáctico, realmente. 

Normally the direct/inverse operator monitors the direction of source -> destination in the hierarchy of referents, but there are exceptions. In verbs of involuntary perception like 'see' it monitors undergoer -> source, since source is not a real actor here.
Así pues, por ejemplo:


yo te veo -> ver(an,sg).dir-nv yo.dst tu.src (direct marking even though src->dst is 2>1)

yo te miro -> ver(an,sg).dir-vol yo.src tu.dst

yo veo montañas -> ver(in,pl).pat yo.dst montañas.src

yo miro montañas -> ver(in,pl).agt yo.src montañas.dst

tú me das miedo -> tu.src fearful(an,sg)-inv.nv yo.dst

los truenos me dan miedo -> trueno.src fearful(an,sg).pat yo.dst

yo corro -> yo.curr correr(an,sg).agt

yo te doy galletas -> dar(an,sg) galletas.curr tu.dst yo.src

yo cojo hierbas -> coger(in,pl).agt yo.src hierbas.dst

Depende de la clase del verbo, obviamente.

Controlled predicates select actor (normally source) as the most salient argument.
Uncontrolled predicates select undergoer (normally destination).

Clause chaining monitors this salient argument.

So:

Cogí hierbas y corrí (C)

coger(in,pl)-pfv1-agt-same 1-correr 

el trueno me asustó y huí (NC)

trueno.src miedoso(an,sg).pfv1-pat-same 1-huir

tú me asustaste sin querer y huí (NC)

miedoso(an,sg)-pfv1-inv.nv-2-same 1-huir.agt

tú me asustaste y huí (C)

1-miedoso(an,sg)-pfv1-inv.v-2-diff 1-huir.agt

me asustaste y te escondiste (C)

1-miedoso(an,sg)-pfv1-inv.v-same 2-refl-esconder.agt

me asustaste y te gritaron (C-C)

1-miedoso(an,sg)-pfv1-inv.v-2-diff 2-gritar(an,sg)-inv.v

me asustaste y te asustaron (C-UC)

1-miedoso(an,sg)-pfv1-inv.v-same 2-miedoso(an,sg)-inv.nv

So it checks the most salient of one predicate against the most salient of the next. Intransitive predicates obviously select their only core argument.

Me parece bien ^^

Thing is, actor > undergoer > topic for the direct-inverse marking. We've said, in 'see' there's no real actor. The thing seen, even though it gets marked as source, is semantically a topic. It doesn't have a role in the action. 

With 'look' we have actor > topic then... OK. With 'scare', it's an (involuntary) actor > undergoer. And 'know'? Subject is an experiencer... so undergoer too. Destination case. What about the matter known? 

Gah, it works like 'see'. It's a mental event, but perception nonetheless. 

'I desire you' -> 'you make me horny (not something you control)'

2.src 1.dst horny-inv.nv

'you seduce me' -> 'you make me horny (deliberately)'

2.src 1.dst 1-horny(an,sg)-inv.v-2 XD

What about themes getting marked by current case?

So 'I see you' I.dst see(an,sg)-dir.nv you.curr

And 'I look at you' I.dst see(an,sg)-dir.v you.curr

Cases are more of a semantic notion... but syntax is not without influence. The argument of an intransitive verb may not be either source or destination, only locative and current. 

In fact, in intransitive clauses semantics are more a verbal issue, cases are more syntactically determined. In transitive clauses, it's a little different. 

'I know history'

I is syntactically A and semantically undergoer. History is syntactically P and semantically theme. Therefore, it is a non-controlled transitive predicate with inanimate P. This implies:
- A gets marked as destination
- P gets marked as current
- The control morpheme inflects only for the semantic status of the A: patient
- The root inflects for the gender and number of P: inanimate and default number  

1-know(in)-pat history.cur

Control morpheme also reflects the direction of A > P in the hierarchy 1 > 2 > 3 when both A and P are animate. 

This is fascinating!!!

It's not clear whether sentences like 'thunder scares me' are transitive or intransitive, since inanimate A's leave no trace on the verb and the gender inflection is the same for P and S. 

Overall:

Case-marking schema depends thus on the semantic class of verb. Like this:

Intransitive:
 Motion/Position -> S: Current eg run            
 Elsewhere -> S: Locative eg quiet
Transitive:
 Agent-patient -> A: Source - P: Destination eg eat
 Agent-theme -> A: Source - P: Current eg look
 Patient-theme -> A: Destination - P: Current eg see, know
Ditransitive:
 Source-Current-Destination eg give

The control morpheme reflects 
a) Controlled vs non-controlled predicate
b) The flow of direction between animate A and animate P according to hierarchy 1 > 2 > 3

Stems inflect for gender and number of P or S arguments. 

Causative of intransitive -> Agent-patient transitive
Causative of agent-patient transitive/ditransitive -> Causative subject is A, causee gets instrumental.
Causative of agent-theme -> Not allowed
Causative of patient-theme -> Becomes normal ditransitive.

So 'I made him write a letter' is essentially the same structure as 'I wrote a letter with a pen'.

NICE!

Partitive overlaps with the other cases. Difference between habitual and progressive lies here: progressive actions get a ptv-marked P. 

Order's syntactic. A and P position have no tie with semantic role. 

Ahora vienen las pasivas y antipasivas... las morfológicas eliminan totalmente el argumento correspondiente manteniendo la transitividad del verbo.

Luego estarían las alcanzadas mediante tópico. Éstas son más débiles. 

Y luego tenemos lo del clause-chaining. ¿Realmente hacen falta? Se podrían enganchar oraciones sin más y ya. Y eso que me ahorro. 

En realidad no cuesta mucho... para que algo sea seguido por el clause-chaining, debe ser animado y definido. Si no lo es, pasivas y antipasivas según convenga, teniendo en cuenta que los controlados rigen agente y los no-controlados rigen paciente.

Vale. Me parece bien. 

Y la correferencia en subordinadas? Supongo que podría ser un 'pronombre'. ¿En base a qué? Acusativo no? Posiciones A y S son las que cuentan. El pronombre de 4ª por así llamarlo aparece en subordinadas y posesivos, básicamente. 

En teoría en clause-chaining el aspecto marca la simultaneidad o secuencia de las acciones... nada más.

Los verbos se pueden componer si comparten argumentos y morfología. Si sólo comparten oblicuos, -te. Si no, clause-chaining. Si son dos cosas más independientes, ya pasamos a subordinadas. No? Esto ya lo veré a la hora de traducir. 

En realidad -te es una subordinada no? El scope de los oblicuos viene dado por el orden... no, pues para -te deben compartir ^^

El compounding y el clause-chaining cumplen mucho de adverbios y conjunciones. 

Si una entidad no es específica, la distinción por número desaparece.

wolfsongalaska.org

http://forum.wordreference.com/archive/index.php/t-210859.htm

I aspire no higher. I desire no better.