Relative Clauses

[Yehuda N. Falk]

Thanks to Paul Kroeger, Miriam Butt, Helge Dyvik, and Stefan Mueller for
sharing thoughts on relative clauses. I'll have to get my hands on the
Grammar Writer's Cookbook -- I knew that there was something I was going to
end up regretting not buying at CSLI when I was at Stanford :(

Helge and Stefan argue for an anaphoric connection. I am still uneasy about
this for English for the following reasons:

-Where does the [PRED 'PRO'] feature come from? If there's a wh relative
pronoun, it would be the source of the feature. But what about relatives
without the relative pronoun? Can PRED values be introduced
constructionally? And if the relative clause construction can introduce a
[PRED 'PRO'] wouldn't that always block the relative pronoun (by Economy of
Expression). It *does* block them in infinitival relatives; cf. "a chair to
sit on" with "a chair on which to sit" but "*a chair which to sit on." But
in finite relatives the wh isn't blocked. And this bothers me, and makes me
think that things are more complicated than they look.

-Related to the above: to my deranged mind, at least, it seems more natural
for the relative clause to be annotated "down TOPIC = up" rather than with
a TOPIC's PRED feature. Now I agree that it can't be quite as simple as
"down TOPIC = up" (that is, the relation isn't complete identity), but I
wonder if it's partial identity. In my original post, I suggested a shared
PRED feature. Another possibility might be to relate it to the DEF feature;
there was an idea in transformational work--I believe late 60s-early 70s,
but I have no idea who--that the relative clause is atached to the
determiner in deep structure. (Jackendoff 1977 X-Bar Syntax suggests a
close semantic relation between the relative clause and the determiner,
although he rejects the transformational approach.)

-I wonder how to distinguish between restrictive and nonrestrictive
relative clauses in f-structure. Nonrestrictives require the wh; this would
be natural if [PRED 'PRO'] comes from the wh pronoun. If all relatives
involve an anaphoric link, why is the wh obligatory in nonrestrictives but
not in restrictives?

One thing we have to remember is that "relative clause" isn't an analysis
-- it's a taxonomic classification. Different languages may use different
formal devices to achieve the same consequence, and even one language may
use multiple devices. Consider "obligatory equi", which is basically a
semantic property of the governing verb: LFG posits distinct formal ways
that a language may achieve the desired result: functional control,
anaphoric control, and a complex predicate construction. It is possible
that, for example, wh relatives achieve the relation to the larger NP one
way and non-wh relatives do it a different way. This dissociation between
functional constructions and formal devices is one of LFG's strengths in
dealing with cross-linguistic variation.



On a slightly different note, Paul raised a question concerning resumptive
pronouns: whether their link to the relative TOPIC is anaphoric rather than
identity. I think that has been assumed; for example, in Joan's forthcoming
book. Here again, I think things are more complicated; generally (though
not completely), relativization-by-gap blocks relativization-by-resumptive
pronoun, which would seem to imply that the f-structures are the same (so
Economy will choose the one with fewer nodes; i.e. no resumptive pronoun).

--Yehuda


                             Yehuda N. Falk
        Department of English, The Hebrew University of Jerusalem
                      Mt. Scopus, Jerusalem, Israel
                         msyfalk at mscc.huji.ac.il
       Personal Web Site    http://pluto.mscc.huji.ac.il/~msyfalk/
      Departmental Web Site    http://atar.mscc.huji.ac.il/~english/

"And because,  in all the galaxy, they had found  nothing more precious than
Mind,  they  encouraged  its dawning everywhere.  They became farmers in the
fields of stars; they sowed, and sometimes they reaped." --Arthur C. Clarke,
2001: A Space Odyssey