Re: nom/accu pronouns erg/abs everything else

Jeff Rollin wrote:
> In the last episode, on Mon, 14 May 2007 11:52:58 -0400, Tim Smith
> <tim.langsmith@...> wrote:
>
>> Reilly Schlaier wrote:
>>> my conlang just morphed in this weird thing (to me)
>>> i've got erg/abs alignment everywhere except in my pronouns
>>> anyone ever heard of that
>>> cause i havent
>>> and it seems very odd
>>>
>>>
>> This is actually a fairly common pattern.  IIRC, Guugu Yimidhirr (and
>> many other Australian languages) have nominative-accusative alignment
>> for all pronouns (or at least all personal pronouns) and
>> ergative-absolutive alignment for all nouns.
>>
>> Interestingly enough, it never works the other way; there are, AFAIK,
>> no known natlangs with nominative-accusative alignment for nouns and
>> ergative-absolutive alignment for pronouns.
>>
>> This is actually a specific instance of a more general principle.
>> There's an animacy hierarchy that seems to be more or less universal
>> cross-linguistically, that runs roughly as follows:
>>
>> 1st and 2nd person pronouns > 3rd person pronouns > animate nouns >
>> inanimate nouns
>>
>> If a given language has more than one alignment pattern, it's always
>> nom.-acc. for nominals higher on the hierarchy and erg.-abs. for
>> those lower, never the other way around.  The cutoff point can be
>> anywhere on the hierarchy.  There can even be two cutoff points, one
>> (point A) above which everything is nom.-acc. and another (point B)
>> below which everything is erg.-abs.  If point A is above point B, you
>> get an area in between where neither alignment applies (thus no case
>> marking, or at least none for acc. or erg.); if point A is below
>> point B, you get an area of overlap in which there's a tripartite
>> system.  I think all of these possibilities exist somewhere among the
>> Australian languages.
>>
>> - Tim (momentarily de-lurking)
>
> This is interesting. Are there any languages which display a split
> nom-acc/tripartite system, e.g.
>
> Pronouns = Nom/Acc
> Everything Else = Erg/Acc/Nom
>
> or
>
> 1 & 2P = Nom/Acc
> 3P & EE = Erg/Acc/Nom?
>
None that I specifically know of, but I wouldn't be surprised if such a
language existed.  It would not be incompatible in any way with the
general rule I've summarized.

In retrospect, I spoke carelessly when I said "I think all of these
possibilities exist somewhere among the Australian languages."  I really
don't have enough data to back up such a generalization.  What I should
have said is that, based on what I do know, I know of no reason why any
or all of these possibilities should not exist, and that the most likely
place to look for them would probably be among the Australian languages,
which seem to specialize in unusual case systems.

By the way, Paul Bennett's recent post, quoting Dixon extensively, says
essentially the same thing I did, but he says it in more detail, and
includes a plausible explanation for _why_ this generalization should hold.

- Tim


> Jeff
>