Oil In My Lamp wrote: ↑28 Mar 2020 16:01
Close, but not quite. If the verb is genuinely intransitive - rather than merely univalent - then it cannot have a patient.
In true 'ergative-absolutive' languages, the subject of an intransitive is always treated the same way, regardless of its semantics.
So, does that mean "I ate" is univalent rather than truely intransitive?
I ate - univalent
I slept - intransitive
We shouldn't get too hung up on terminology, because it isn't set in stone. But yes, I would say that for most purposes "I ate" is a transitive but univalent verb, which therefore has an implied but unstated semantic patient (the thing you ate).
However, it's more debateable whether this is always the case. You could argue that in an exchange like "do you want lunch?" - "no thanks, I already ate", the verb "to eat" is acting as a true intransitive, with a meaning like "I partook of a meal". Our verb "to dine" is (arguably) truly intransitive in this way, and cannot be transitive). In particular, some might say that if someone "ate at the Ritz", they didn't even necessarily eat anything at all (maybe they just had water, but the person they were with ate) - that's probably true of 'dine', but maybe not of 'ate'. Semantics are always wobbly.
But yes, you seem to grasp the distinction I'm making. However, note that BOTH verbs there are syntactically univalent, even though the first is semantically transitive.
And so is "I" in "I ate" ergative? Or it could be?
Well, English doesn't have an ergative.
In an ergative-absolutive language, could you have a sentence like "I ate", with "I" in the ergative case? Yes. But I'm guessing it's not that common.
There's two ways a language can treat a sentence like "I ate". One is as a bivalent verb with a non-overt patient. In an erg-abs language, the patient is the primary argument of a transitive, so it would be very strange to simply drop it like that - it would be like saying "ate bread" in English, with no subject. Languages usually only allow this when the missing argument is present close by (eg in an adjacent clause), to the extent that a linguist would probably say it isn't missing. Some languages are 'null subject' languages, which can drop the subject - but this is treated as underlyingly having an invisible pronoun as the subject, following the usual rules to determine what the pronoun refers to, and it also usually only happens when the verb agrees with properties of this subject, so the subject isn't really missing at all. Some languages also have situations where a null subject really is null, but these are usually restricted to a very small number of verbs, such as weather verbs (where English has "it is raining", some languages just say "is raining").
I would expect something similar with patients in ergative languages, though not necessarily (as most ergative languages are only partly ergative - so I assume some (many?) act like nom-acc languages with regard to dropping patients?)
IF an erg-abs language allows patient-dropping in this way, then "I" would indeed be in the ergative there.
The other way, however, to treat "I ate" is as a univalent verb with no patient at all (and just ignore the fact that, semantically, there COULD be a patient). This could be done lexically - the way that English has both univalent/bivalent "I ate (the bread)" and strictly univalent "I dined". It could also be done inflectionall, the way that English has bivalent "I ate the bread" and univalent "the bread was eaten". If a language does this, then no, "I" would not be in the ergative, it would be in the absolutive, because it's now the subject of an intransitive verb.
I-ERG ate bread-ABS
An inflexional process that takes transitive verbs and yields intransitive verbs by removing the patient is called an "antipassive voice".
And of course, for extra fun, an antipassive could be zero-derived, with the valency determined only from word order, case marking and so forth.
Salmoneus wrote: ↑28 Mar 2020 01:19
However, there are forms of "split ergativity" - most typically the "active" alignments found in some native american languages - where patient-like subjects and agent-like subjects are indeed treated differently (basically, S=A (nom-acc) when S is semantically sufficiently agentlike, but S=O (erg-abs) when S is semantically sufficiently patientlike).
That sounds like an easy way to do it, but maybe it would get very confusing after a while depending on social expectations in correlation to how lax the grammar rules are.
Grammar rules are never lax, although they may vary with dialect.
Split ergativity is no more confusing than any other alignment. It can be divided into two types: fluid-S, in which the marking of the sole argument of univalent verbs depends upon the actual circumstances of the event; and split-S, in which the marking of the argument is fixed for each verb.
EDIT: I will say, though, that I'm not a linguist, so I'm welcome to being corrected by those that know what they're talking about. However, I think the above is fairly true, on the basis of everything I've read over the years.