Linguistic commentary from a guy who takes things too literally

Subject and Object Gaps in Coordinated Relative Clauses

Posted by Neal on September 18, 2006

Last year, I wrote about reading Doug a book I’d read when I was his age and held onto all these years. At that time I found an unusual usage of the word enjoy where I’d have expected suffer from. Now I’m reading the same book aloud to Adam, with Doug listening in and keeping his spoilers to himself. And what do you know–I’ve found another linguistically interesting quotation in the book. Here it is:

In the another corner they’d found a creeper, which [Tubby had left behind], and [was particularly helpful].
(Clifford B. Hicks, Alvin’s Swap Shop, p. 32)

It’s another coordination of relative clauses in which one is missing an object (Tubby had left    ) and one is missing a subject (    was particularly helpful). Arnold Zwicky wrote about coordinations like this last year, and offered the following examples (for attributions, see Zwicky’s post):

…New Mexico, which [the president leads] but [was still uncalled as of noon Wednesday]

So for people who [I’m not going to give a cox-2] and [also have a history of ulcers], the way around it is to take the anti-inflammatory and make it into a cream.

Reading that post sensitized me to this one that I heard and wrote about a little later:

…and the kids [I don’t actually serve] but [are here today].

And since then, every now and then Arnold will send an email to me and others who find this kind of thing interesting with more examples he’s come across, such as:

something that [President Bush supports] but [has been rejected by the House] (Carl Kassell)

but maybe this is a construction that [writing teachers have noted and corrected in student writing for decades], but [never got turned into an explicit rule in the advice literature]. (Zwicky himself)

Something that [both Foerster & Steadman and Kierzek get more or less right], but [tends to be downplayed in later advice about the passive], is that how “important” the referent of Y is] plays a role in choosing the voice for a clause. (Zwicky himself again)

Not all of these examples go over equally well; in particular, the cox-2 example was difficult for Bruno Estigarribia, who was the one who brought it to Zwicky’s attention. But to the extent that these coordinations of object-gap and subject-gap relative clauses are all at least grammatical enough to be generated by (I assume) native speakers of English, I can note two patterns:

  1. The first relative clause always has an object gap, and the last one always has a subject gap. I don’t know if there’s any significance to that yet.

  2. Two: In all but the troublesome cox-2 example, the subject gap is in a clause whose subject doesn’t have strong, typically subject-associated properties such as being the actor or causer.

This second property is worth looking into. Three of the subject-gap clauses are true passives, where the subject is a patient (was still uncalled, has been rejected, never got turned into), and one is as good as passive, with a passive infinitive following a raising verb (i.e., one whose subject gets all its properties from the verb that it’s paired with, as happens in tends to be downplayed). Two more have forms of be for their verbs (was particularly helpful, are here), and are not used to ascribe agent-like properties to their subjects. So it looks like there’s a semantic component in the explanation for why some of these coordinations are better than others: If the item described by the relative clause fills a patient(-like) role for each coordinated clause, the coordination is more likely to be OK, even if the patient role corresponds to an object in one clause and a subject in another.

This makes me wonder: Could I also construct a mixed subject/object gap relative clause coordination where the described item fills an agent-like role in each coordinated clause? It’d be tough. There are plenty of verbs whose subjects aren’t agents (e.g., die, suffer, like), but it’s harder to think of verbs whose direct objects are agents. I think the best bet is getting a causative verb like walk, whose direct object is still somewhat agent-like, being animate and having some control over its actions. So let’s see…

Fifi is the dog that [the dogsitter walked] and [bit the other dogs].

Nope, no good for me. The only way I can parse it is with the dogsitter biting the other dogs after walking Fifi. (Kind of like words you look back on and cringe.) Any of you care to try creating a more grammatical example?

3 Responses to “Subject and Object Gaps in Coordinated Relative Clauses”

  1. Glen said

    All of those coordinations bother me, though some less than others. I would fix them by inserting a second “that” or “which” before the second relative clause.

  2. Russell said

    Well, there are verbs like “dress” and “feed,” whose patients are not really agents, even in the subevents of eating and putting-on (I suppose the walker in transitive “walk” is more construable as an agent):

    that’s the child i (tried to) dress (with nice clothes) but kicked me in the shin.
    that’s the baby i (tried to) feed (dessert to) but just spit it back out.

    I reserve judgement.

    And object-of-preposition gaps in the first clause?

    ralph is the mafia member i exacted revenge on and for whatever reason never tried to get back at me

  3. Szwagier said

    It seems to me that a second ‘that’ in your last example solves the problem completely. If this is true, then the question is about ‘that’-deletion. Is that any help?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: