July 03, 2004

Pronouns with following antecedents in subordinate clauses

The word "antecedent" suggests (Latin ante being a preposition meaning "before") that an occurrence of a pronoun must be linked to a noun phrase (NP) that has occurred before the pronoun occurrence. Not true. The so-called antecedent can come later if the pronoun is in a subordinate clause: When he had finished eating, John got up from the table. There he had finished eating is a subordinate clause and John is in the main clause, and we can certainly understand the pronoun as having John as its antecedent. But can an antecedent come later if the tables are turned? Can you have a pronoun in the main clause coming earlier than an antecedent in a subordinate clause?

The standard view seems to be no. Pronouns can refer back to earlier sentences, and subordinate clause pronouns can refer forward to main clause antecedents yet to come, but main clause pronouns can't refer forward to subordinate clauses yet to come. But the standard view may not be quite right, at least not for pronouns functioning as genitive NP determiners. The following example occurs under the headline "Still taking on the world?" as the subtitle of the first leader in the July 3, 2004, issue of The Economist:

His foreign critics need to notice that George Bush has now done what they want

Since nothing precedes it on the page or in the leader section, the only reasonable analysis is that the pronoun his, which functions as the determiner in the subject NP of the main clause, has as its antecedent a following NP, the subject of a content clause (George Bush has now done what they want) which is the complement of the verb notice, which is in turn head of the subordinate infinitival complement of need, hence is doubly embedded, in the second subordinate clause down. Not every theory of pronominal anaphora predicts this possibility. (The Cambridge Grammar of the English Language cites, on page 1479, the sentence ?Her husband had supported Ann throughout the ordeal, the question mark marking it as being of questionable acceptability. But it doesn't have the antecedent in a subordinate clause. That possibility is not mentioned, either to say it is possible or that it is not. Note, however, the remarks about "Anticipatory anaphora for rhetorical effect" on page 1480, which is very probably what The Economist's subhead is a case of.)

Posted by Geoffrey K. Pullum at July 3, 2004 04:41 PM

It does seem to be a focus/topic thing:

His critics noticed {when | that} John finished eating.
John's critics noticed {when | that} he finished eating.

Both sound all right, but pronouns are assigned reference pragmatically, and the cataphoric 'his' feels as if it would prefer some prior antecedent. Finding none, it can link to a following one, but at a cost of bleaching the reference and focusing on the verb phrase it's in.

Posted by: entangledbank at July 3, 2004 07:40 PM

I found this document (http://www.ecotton.com/newsletters/01022003.pdf) which contains the following passage:

"Before I left on this last trip, Jack and I sat down and 'discussed' our use of terminology in the current software. [...] I told Jack that maybe some of these should be changed to show the merger of the two programs. After he listened to my argument, _he did what Jack always does_. He asked 'why?'"

I believe this illustrates the construction in question. For me, this is marginally grammatical.

Posted by: Russell at July 4, 2004 08:42 AM

The classic -- as in 1960s -- formulation of the constraint is the precede-command constraint: A pronoun cannot both precede and command its antecedent. Terminological and theoretical matters aside, what the constraint was supposed to do is rule out cataphoric subject pronouns, as in
*She knows that Mary is admired. (1)
'Mary knows that she is admired.'
while allowing cataphoric pronouns *within* a subject:
Her mother admires Mary. (2)
'Mary's mother admires Mary.'
Everyone who knows her admires Mary. (3)
'Every who knows Mary admires Mary.'

Russell's example in his 7/4 comment,
He did what Jack always does. (4)
'Jack did what Jack always does.'
is of the same form as (1), and without the context that Russell provides, is just as bad. The context is crucial. The constraint is not entirely a structural one: A pronoun cannot establish a discourse referent that is referred to by a nonpronominal NP that the pronoun precedes and commands. Even this version might have a completely pragmatic account, based on the special pragmatic statuses of subjects.

In any case, like "entangledbank", I think that what's going on here has largely to do with the discourse status of entities and the way in which NPs of various types establish, refresh, and refer to discourse referents.

A few more remarks:

(A) A number of people have pointed out that cataphora is rarely used in discourse-initial sentences; most of the time, a discourse referent has already been established. Still, things like (2) and (3) are possible as discourse-initial sentences.

(B) In fact, sentences like (2) are often actually *recommended* by some usage manuals, as fixes for violations of the (putative) Possessive Antecedent Proscription, as in
Mary's mother admires her. (5)

(C) Despite (B), the usage manuals uniformly recommend caution with cataphora in general, on the (very reasonable) grounds that it can make more work for the reader or hearer and can, as a result, impede understanding.

Posted by: Arnold M. Zwicky at July 5, 2004 12:50 PM

My response to Zwicky's comments, which focused solely the linguistic discussion of co-reference (binding, linking, etc.), which is the topic of this thread. has been deleted.

To talk about the topic of a post is offtopic too?

Posted by: Tony Marmo at July 6, 2004 10:50 PM

And my comments on Pullum's post, which were just about the same linguistic facts, are gone too!

Posted by: Tony Marmo at July 6, 2004 10:51 PM