On May 1, 2010, at 5:22 PM, Alex Fink wrote:

> On Wed, 28 Apr 2010 12:38:16 -0500, Eric Christopherson <[log in to unmask]>
> wrote:
>> I'm working on a diachronic conlang, the parent of which is ergative (in
> some respects at least). I just reread the chapter on grammatical relations
> in _Describing morphosyntax_, and was reminded of two things:
>> 1. All "ergative" languages are split ergative in some respect.
>> 2. In split ergative languages, agreement markers are most likely to show
> nominative-accusative alignment.

I just realized I worded this in a way that too-broadly refers to all split ergative systems; the actual quote in _DM_ is "[W]hen split ergativity is based on the semantic and/or pragmatic characteristics of the noun phrases, it will always be the case that noun phrases higher (to the left) on the topic-worthiness hierarchy will trigger a nominative/accusative subsystem, while noun phrases lower (to the right) on the hierarchy will trigger an ergative/absolutive subsystem."

The hierarchy in question:

1 > 2 > 3 (agreement) >
1 > 2 > 3 (pronouns) >
proper names >
humans >
non-human animates >

-- with "definite > indefinite" running parallel to the rest of the hierarchy.

Now that I've read that again, I'm a little confused as to how noun phrases higher on the hierarchy trigger a nom/acc subsystem, etc. -- what if a sentence has *both* a noun phrase high on the hierarchy and one low on it? -- Does he mean that an NP high on the hierarchy, but in a P/patient/object/undergoer role, triggers nom/acc? And that an NP low on the hierarchy, but in an A/agent/actor role, triggers erg/abs?

(Also I'm unsure of how he can speak of "noun phrases" when the very top of the hierarchy is agreement markers and pronouns. Obviously those can refer to NPs, but I don't know if it's normal to *call* them NPs themselves.)

>> This is troubling: the protolanguage doesn't have agreement markers. I've
> decided to attach pronouns to verbs and have them evolve into agreement
> markers. The problem is that traditionally I have wanted to attach just the
> transitive agent (i.e. ergative) pronouns; so agreement would be shown only
> on transitive verbs. Is that unrealistic?
> Checking Dixon's _Ergativity_ (which I don't really know my way around),
> there is an example of this attested, but it has other unusual features of
> note, suggesting to me that this pattern might have a hard time arising
> directly (without help from sound change, e.g.) from a more "normal" ground
> state.  Grist for the mill here in any case.
> The Nilotic language Päri at least has the agreement pattern which this
> would result in, A being marked on the verb "in one type of construction"
> but S/O never.  That construction appears to be when the A is fronted for
> topicalisation; the unmarked orders are SV, OVA.  
> That's about all that's said on the point of agreement.  But Päri also has
> an interesting split in case marking: in most types of clauses, it's
> ergative with the S/O unmarked and the A carrying a suffix //-i_L//, but in
> imperatives and some subordinate clauses it's a (typologically rare) marked
> nominative system, i.e. the S takes //-i_L// too.  Comparative evidence and
> general trends in which clause types retain archaisms suggest that the
> marked nominative is the older system; further, the original word order may
> have been VS, VAO.   T. Andersen suggests that the ergative may have
> originated in the context of a habitual (pragmatic?) fronting of S and O
> that changed the basic word order: when S was fronted it lost its marked
> nominative suffix, becoming zero-marked like O.  I wonder whether something
> similar mightn't have happened, or happen, to an earlier nom-acc agreement
> system as well.  

That's pretty interesting. I tend not to consider different clause types, unfortunately.

>> Also, assuming that system is somewhat plausible, would it be realistic to
> say that the primordial ergative pronouns would actually have the ergative
> *case marker* on them? And that the case marker would leave some trace on
> the eventual agreement markers?
> I don't see why not.  What rationale would there be for dropping it?

Well, that's one thing I'm thinking about. If language generally is so intolerant of a phenomenon like ergative pronominal or agreement marking, I have no idea what the actual diachronic pathways would be -- would such an agreement system just never spring into being, or would it exist for a time but give way to something more typologically normal?

In relation to this conlang, this means that I'm not sure the ergative marker would ever have been attached to pronouns in the first place.

>> Alas, the specification of the protolanguage is so open that it merely
> lists the ergative marker as a morpheme, but doesn't describe at all where
> it is used.
> That's unfortunate.  What's the protolanguage?  I suspect from your word
> choice it's also a conlang.  (In which case, cool, I love me some
> collaborative diachronic projects.)  Can you not press its creator for better?

It is a conlang indeed. The creator of the protolang wanted to leave things pretty vague to allow more freedom. That's cool in a way, but I get lost in details like this one. There is only one (semi-?)canonical protolanguage phrase with the ergative marker, which translates as "the fingers*.ERG of the cat". I take from that the idea that inanimate objects -- or at least body parts -- can occur in the ergative sometimes; perhaps it's when they are quasi-agentive, as in "The fingers of the cat scratched me" or "The fingers of the cat held a tiny mouse". (I wonder if the ergative might be more of an instrumental here.)

* It is indeed glossed "fingers" and not "claws" or "front toes" :)