Davush wrote: ↑12 Dec 2020 13:51
I have realised in Hakuan the following "problem":
When both agent and patient are 3rd person, and one is pronominal, it is impossible to retrieve which one is the A and which the P. Although both arguments are marked on the verb, they don't have separate A/P forms to distinguish. For example:
saps-oa tupu
see-3sg.3sg man
'the man saw him/her'
OR
's/he saw the man'
Would (or is) this type of ambiguity tolerated in any natlangs?
Yes. English is one example. In English, it typically arises when both arguments are purely pronominal:
"he saw him" - do I mean that he saw him, or that he saw him?
However, it also arises in certain topical and focusing constructions with a full nominal argument that is reiterated with a pronoun:
"as for the man, he saw him" - the man is almost certainly the semantic agent or patient of the verb here... but which?
More generally: yes, these ambiguities can arise in languages even where one argument is nominal.
What type of strategies might arise to disambiguate, without the need for case?
A good question. When a verb has two third-person arguments, how can you tell which is subject and which is object?
The first solution would be: case. Remember that case need not mean morphological case. You could easily adopt a rule where subjects preceded the verb and objects followed it, for instance. Or you could have, say, VSO order. How would that help where one argument was dropped? Simple: don't let the argument be dropped. Either in general, or just for disambiguation.
The second solution would be: better indexing. Your pronouns, or pronominal affixes, could more precisely indicate their reference - they could indicate the gender of their referent, for instance, or relative status.
The third and much broader solution is to look at context. There will usually be one "most likely interpretation" (let's call this MLI for short), and one "less likely interpretation" (LLI). The question then becomes: given that the MLI will occur most often, how do I spot when LLI is intended?
The easiest way is: context! The LLI is rarely intended. If it is, it'll often be in an unusual context where the possibility is very obvious, and the speaker may even overtly flag the oddness of the situation. People can usually spot this from context.
But if you do want an overt LLI marking, you then have a further question: how do you define it? That is, how do you define the MLI (the LLI then being the absence of MLI)? Different languages do this differently.
In essence, languages make assumptions about what the subject is likely to be. The most straightforward assumption would be: the subject of the previous clause will be the subject of this one. A more indirect assumption would be: the topic of the previous clause will be the subject of this one. Or, indeed: the topic of this clause will be its subject. Or even more generally: a high-saliency referent will be the subject. Or: a high-agency referent will be the subject.
In each case, you then have to mark (assuming you're not leaving it to context) clauses where the MLI is not correct. You can do this by:
- marking the noun(s), either morphologically or syntactically. You can mark case, topic or focus; you can also make a proximate/obviate distinction, which basically is directly specifying the saliency of the noun
- marking the verb. When you're indicated a switch in the subject, this is called switch reference. When you're indicating that the subject is actually lower-agency than the object, this is called direct/inverse marking. It's worth mentioning here that verbs don't have to be marked with affixes. You could also, for instance, use a serial verb construction incorporating a univalent verb, to force the interpretation you want.
- upgrading or downgrading an argument. There's a hierarchy of styles of reference: elided subjects > pronouns > nouns. You could have a rule like: unexpected subjects cannot be pronouns, but must be full nouns. Or you could have a rule like: expected subjects are elided. Likewise with agreement on verbs.
---------------
Some concrete examples? Let's assume we're trying to disambiguate "a hunter shot the bear" and "the bear shot a hunter". This language is VSO, and both arguments must be indexed on the verb. Arguments can, we'll say for now, be elided. So, our basic form is: "shot-he-he hunter". How do we tell if the hunter is shooting, or being shot? Some ways:
Fronting/topicalisation of unexpected argument:
enter-33 hunter wood. shot-33 bear. - the hunter shot the bear.
enter-33 hunter wood. bear shot-33 . - the bear shot the hunter.
Topicalisation:
hunter enter-33 wood. shot-33 bear. - the hunter shot the bear. [topic is subject]
bear enter-33 hunter wood. shot-33 - the bear shot the hunter. [topic is subject; the bear is the topic of both clauses, despite not being an argument in the first]
Obviation:
enter-33 hunter wood. shot-33 some bear. - the hunter shot the bear. ['some' marks an obviate]
enter-33 some hunter wood. shot-33 bear. - the bear shot the hunter.
Proximal marking overriding definiteness:
enter-33 a certain hunter wood. shot-33 the bear. - the hunter shot the bear.
enter-33 a hunter wood. shot-33 the bear. - the bear shot the hunter.
[the bear is assumed to be subject because it's definite and hunter is indefinite. But the hunter can be promoted above bear with special marker, here 'certain']
Switch reference:
enter-33 hunter wood. shot-33 bear. - the hunter shot the bear.
enter-33 hunter wood. but shot-33 bear. - the bear shot the hunter. [well, or 'the wood shot the bear', but context...]
Direct/inverse:
enter-33 hunter wood. shot-33 bear. - the hunter shot the bear.
enter-33 hunter wood. shot-33-wow bear. - the bear shot the hunter. [human hunters outrank bears, hence the low-ranked bear shooting a hunter must be marked.]
[Direct/inverse implies a fixed hierarchy, usually not very precise. This example could also instead be admirativity: just marking any pragmatically surprising news]
And then, going back to same-reference as an exemplar assumption, to show other methods for encoding LLI (could also be used with other assumptions, such as with direct/inverse...)
Switch reference by elision of expected subject pronoun:
enter-33 hunter wood. shot-33 bear. - the hunter shot the bear. [pronoun dropped because expected subject]
enter-33 hunter wood. shot-33 bear it. - the bear shot the hunter.
Switch reference by elision of expected subject agreement AND pronoun:
enter-33 hunter wood. shot-3 bear. - the hunter shot the bear. [pronoun and agreement dropped]
enter-33 hunter wood. shot-33 bear. - the bear shot the hunter.
Switch reference by non-pronominalisation/non-elision of unexpected subject:
enter-33 hunter wood. shot-33 bear. - the hunter shot the bear. [subject elided leaving only agreement]
enter-33 hunter wood. shot-33 bear hunter. - the bear shot the hunter. [unexpected subject cannot be elided]
Switch reference by intransitive verb:
enter-33 hunter wood. shot-33 bear. - the hunter shot the bear.
enter-33 hunter wood. come-3 bear shot-33. - the bear shot the hunter. [adding the intransitive verb with 'bear' unambiguously as subject changes the expect subject of following 'shot']
And so on and so forth....
N.B. these rules can be different depending on whether the preceding clause is in a different sentence, and whether any conjunction appears. For instance, in English we can say "the hunter entered the wood and shot the bear", but not (in formal speech) "the hunter entered the wood. Shot the bear." - elision rules are different with and without conjunctions.
Also N.B. sometimes rules can exist for determining OBJECTS, rather than subjects - Kiowa can have switch-reference for objects. But this is much rarer.