Switch-Reference (was: Re: Fluid-S pivot in Old Albic)
|Date:||Tuesday, August 9, 2005, 22:02|
--- In firstname.lastname@example.org, Jörg Rhiemeier <joerg_rhiemeier@W...>
> > > To make things more complex, there are also switch-reference
> > > pronouns. The agentive case of the switch-reference pronoun,_ra_,
> > > is coreferential to a patient in the preceding clause, while its
> > > objective case, _ram_, is coreferential to an agent.
> I am not even sure whether "switch reference" is the correct term
> for it. The switch reference morphemes in Amerindian languages
> mostly have to do with subordinate clauses, it seems, though
> I haven't really understood those switch reference systems yet.
In a previous post I gave some URLs discussing switch reference.
I repeat them below.
First, here is a lot of what LinguaLinks has to say about switch
| Switch reference is a grammatical category with the following
| It signals the identity or nonidentity of the referent of an
|argument of one clause, usually its subject, with an argument of
|another clause, which is likewise usually the subject.
|Switch reference functions to avoid ambiguity of reference; for
|example, it may distinguish between two referents that are third
|person and that, thus, may not be otherwise distinguished on the
|It relates clauses, usually adjacent, that may be subordinate or
|coordinate to one another.
|It is expressed
|usually by inflectional affixes on the verb
|sometimes by the same affixes that express subject-verb agreement
|within the clause, and
|rarely by a morpheme independent of the verb.
| Here are some kinds of switch reference:
| What is a different subject marker?
|What is a same subject marker?
| Switch reference is a kind of
| What is discourse deixis?
|Source: Haiman and Munro 1983 ixxiii
So, clearly, switch-reference can be between clauses that are either
co-ordinated, or subordinated.
If you also follow up on serial-verb constructions, you will see that
switch-reference can compare the current clause with a reference
clause that can be either 1) its matrix (if it is a subordinate
clause) 2) the next conjugand (if it is a coordinately conjoined
clause) 3) the previous conjugand (if it is a coordinately conjoined
clause) 4) the first clause in a clause-chain or 5) the last clause
in a clause-chain.
I think that you could probably put 1) and 4) together, kind of, and
find a natlang where a deeply-embedded subordinate clause -- say, a
subordinate clause within a subordinate clause within a subordinate
clause -- could mark itself for referring (or not) to (some of) the
same argument(s) as the over-arching "mother of all" clauses --
the "matriarch" clause -- the matrix of the matrix of the matrix.
I have asked on the list before if anyone knows of any stats on
whether such clauses, when they do have to refer to an argument of a
superordinate clause other than their own immediate matrix, are more
likely to need to co-refer to one of the arguments of the highest
clause than one of the intermediate ones; but no-one has said
The Web Article on "subordinate clause within subordinate clause
within subordinate clause" would lead me to think such a sentence
should come up about once every 250,000 to 1,000,000 sentences.
There must be about 1,000,000,000,000 sentences on the web, mustn't
there? So shouldn't there be between 1,000,000 and 4,000,000
clause sentences on the web? Of course, I don't know how many of
those will have a deeply-imbedded clause that has to refer to an
argument of an ancestor that is not its own immediate matrix; but,
even if it's only 1 in a thousand, that should leave me with a
statistical sample of 1000 to 4000, shouldn't it? (If it's more like
1 every 250,000 to 1,000,000, that leaves me with a sample size of
about 4, which is probably too small to make any significant
Here are the URLs for switch-reference.
Tom H.C. in MI