pronoun trace

Arnold Zwicky zwicky at CSLI.STANFORD.EDU
Sun Feb 17 17:50:46 UTC 2002


ron butters cited the following NP from a newspaper story:
 >a Fort Bragg soldier whose superiors refuse to let him resign [1]
and asked:
 >Do YOU have an obligatory object deletion in this syntactic
 >environment?
meaning that ron finds [1] ungrammatical and requires instead:
 >a Fort Bragg soldier whose superiors refuse to let resign [2].
several others have agreed with these judgments, and now rudy
troike has suggested an account of [1] in terms of parasitic gaps
(i think he means resumptive pronouns) and long-distance effects.
(the discussion of the case of subjects of non-finite clauses is
entirely independent of the issue of [1] vs. [2].)

i'm puzzled, but fascinated, by this discussion.  for me, [1] is
straightforwardly grammatical, and [2] is just as straightforwardly
ungrammatical.  for me, there is no way in which [1] is some sort of
production error, or some sort of attempt to patch up a
hard-to-interpret structure; long-distance effects, resumptive
pronouns, and parasitic gaps are all irrelevant.  in fact, [1] (and
not [2]) is exactly what you'd expect, given the general structure of
WH relative clauses in english.

i'll elaborate, in terms as non-technical as i can manage.  first,
english has three sorts of restrictive relative clause modifiers,
three sorts of clauses that follow, and modify, a head NP h - WH
relatives, THAT relatives, and zero relatives:
  the dog which I saw
  the dog that I saw
  the dog I saw.
it's only the first of these, WH relatives, that we're interested in
here.

next, WH relative clauses have two parts:
  an initial phrase p, of type X, plus
  the rest of a clause, missing a phrase of type X.
in simple examples, the initial phrase is just a single WH word; this
word can function as the subject of the relative clause (in which case
what follows is just the predicate - a clause without its subject), or
it can serve in some non-subject function (in which case what follows
is a clause with a subject, but missing some other phrase):
  subject relative: which  +  saw me   [in: the dog which saw me]
  direct-object relative: which  + I saw  [in: the dog which I saw]
  preposition-object relative: which  +  I stared at
    [in: the dog which I stared at]
  adverbial relative: where  +  I started
    [in: the place where I started]
  etc.
the WH word in such examples is a pronoun anaphoric to the head NP
h ("the dog" or "the place").

but...there's more!  in WH relatives, the initial phrase doesn't have
to be a single WH word; it can be a multiword phrase *containing*
a WH word.  there are two very common types, involving prepositional
phrases with WH-word objects and NPs with WH-word possessive modifiers
(plus the exciting possibility of both in one package).  here are
examples of (non-subject) relatives of all three sorts:
  PP containing WH-word object:  at which  +  I stared
    [in: the dog at which I stared]
  NP containing WH-word possessor:  whose dog  +  I saw
    [in: the man whose dog I saw]
  PP containing NP object containing WH-word possessor:
    at whose dog  +  I stared  [in: the man at whose dog I stared].
such examples involve two interlocking sets of relationships: the
whole initial phrase p is related to the absence of a phrase in the
rest of the clause; and the WH word *inside* p is a pronoun anaphoric
to the head NP h.  in "the man whose dog I saw", the initial "whose
dog" of "whose dog I saw" corresponds to the direct object of "saw"
that's missing later in the clause, and the "whose" of "whose dog" is
anaphoric to the head NP "the man".

these observations carry over to subject relatives, where the initial
phrase p can contain a WH-word possessor:
  whose dog  +  saw me  [in: the man whose dog saw me].

this is exactly the structure we have in ron's original example:
  whose superiors  +  refuse to let him resign
    [in [1]: a Fort Bragg soldier whose superiors refuse to let him
     resign]
the only complication here is an irrelevant one: the predicate "refuse
to let him resign" contains an ordinary (not WH) pronoun "him" that is
anaphoric, to "whose" (parallel to the anaphoric relationships in "his
superiors refuse to let him resign"), and hence, at one further
remove, to the head NP "a Fort Bragg soldier" as well.  this is just
like
  whose dog  +  saw him  [in: the man whose dog saw him]
  whose dog  +  saw him weep  [in: the man whose dog saw him weep]
  whose dog  +  made him weep  [in: the man whose dog made him weep]
  whose dog  +  tried to make him weep
    [in: the man whose dog tried to make him weep].
note that we do not get:
  *the man whose dog saw 'the man whose dog saw him'

the issue is then about the status of examples like the following:
  ?the man whose dog saw weep
  ?the man whose dog made weep
  ?the man whose dog tried to make weep
  ?a soldier whose superiors refuse to let resign.
these are just awful for me (and were so before i went through the
exercise of analyzing them explicitly).  but perhaps there are
people for whom they're ok; it would be nice to have examples from
corpora of various sorts.  such examples would be interesting because
they do *not* conform to the generalizations about restrictive WH
relative clauses in english, but seem to represent some new
development, presumably an amalgam of the subject WH relative
   a soldier whose superiors refuse to let him resign
and the *non-subject* *zero* relative
   a soldier we refuse to let resign
(where a NP is obligatorily missing: *a soldier we refuse to let him
resign).

arnold (zwicky at csli.stanford.edu), with some interest in
  amalgam subordinate-clause constructions



More information about the Ads-l mailing list