On 4/21/06, Eldin Raigmore <[log in to unmask]> wrote:

> >[snip]
> >AFA I understand from Blake's book this is typical of
> >Algonquinian languages.  The question is if it is *so*
> >typically Algonquinian that it is unrealistic in a
> >non-Algonquinian language?
> I believe there are Meso-American and South-American languages, thousands
> of miles away from Canada, that have these systems, too.

Yup.  (see below)

> >BTW would it be unrealistic for the direct voice to
> >be unmarked?
> Not at all; IMO that woud be _realistic_.  The Direct Voice expresses the
> expected situation (the more animate participant is the Agent, the less
> animate participant is the Patient); the Inverse Voice expresses the
> _unexpected_ situation, so it should be "marked".
> >

Indeed.  It would definitely be unusual for the indirect voice to be
the unmarked one.  I'd shy away from saying it would be *impossible*,
but an unmarked indirect wouldn't be something I'd expect to find.

> >I envisage Kijeb as something of a mixture, with both
> >Hierarchical word order and verb marking, as well as nom/acc
> >marking for animates, as well as Split-S/Fluid-S, and the
> >daughter languages (perhaps not all of them) developing
> >split ergative marking.  Perhaps it is altogether
> >unrealistic, or at least highly redundant, to have it all in
> >the same bag!

Inverse/Hierarchical and Split/Fluid-S, at least, are perfectly
compatible.  Spike Gildea (1998) reconstructs an Inverse/Split-S
system for the Cariban family, similar IIRC to the one that survives
in Hixkaryana.  (For those looking for word-order correlations,
Hixkaryana is OVS regardless of direct/inverse.  Not unusual for a
Cariban language.)  (Oh, and to be clear, what Gildea terms Inverse
might be what others term Hierarchical.)

The combination has always struck me as a priori pretty reasonable. 
You have a set of agent-marking affixes if the agent is the highest in
the hierarchy, and a set of patient-marking affixes if the patient is
highest.  Intransitive verb?  If it's an agent, use the agent affixes;
if it's a patient, use the patient ones.

I don't recall if any of Proto-Cariban's daughters evolved into a
split ergative system, but I recall that some of them have ergative
alignment.  (I can't seem to find Gildea's book tonight.)  I see no
reason that it would be "unnatural" for an Inverse/Split-S agreement
system to evolve into a split-ergative agreement system.  Maybe it
hasn't actually happened before, but it's not a crazy/out-there idea.

-- Pat