October 14, 2007

More gapless relatives

Mark Liberman has just posted about this instance of a gapless relative in (non-standard) English:

How can we provide a service that the consumer goes, "Wow, you really made this easier for me"?

As it turns out, non-standard English has (at least) three types of gapless relatives, two with pronouns instead of gaps, and the type above, with neither a gap nor a pronoun. 

The example above has a NP of the form

a service   that the consumer goes X

(where the head is italicized and the relative clause bolded).  There's no gap in the relative clause corresponding to the head, nor is there any pronoun performing that function.  The head picks out a type of thing (in this case, a service), and the relative clause gives us characterizing details about the particular instance of this type; the example above is roughly paraphrasable as "the sort of service such that the consumer goes X".

This is not an especially convincing example of a gapless relative, since it might be analyzed as merely missing a preposition, that is as a truncated version of

a service   that the consumer goes X about

 (Recall our discussion of missing prepositions in a series of postings that began in 2005 and picked up again this year.)

But it's not hard to find other examples that don't submit so easily to this analysis.  Here, for example, is Peggy Noonan in the Wall Street Journal, as reported to the newsgroup sci.lang by Ron Hardin on 6/29/06 (italicization and bolding as above):

Frank Rich is running around with his antiwar screeds as if it's 1968 and he's an idealist with a beard, as opposed to what he is, a guy who if he pierced his ears gravy would come out.

A version with explicit anaphora to his ears is just as problematic as the original:

a guy who if he pierced his ears gravy would come out of them

The pronouns he and his in the subordinate clause if he pierced his ears aren't relevant here, as you can see from recastings with such that:

a guy such that if he pierced his ears gravy would come out (of them)

(which is clunky but standard English) and with a gapped relative:

a guy who if he pierced his ears ___ would have gravy coming out of them

(which is also standard).

One more, from radio station KFJC's Robert Emmett, on the Norman Bates Memorial Soundtrack Show of 3/12/05:

There are films that you are lucky that you don't have to sit through the whole thing.

I'll call these NoPro gapless relatives, to distinguish them from gapless relatives with "resumptive pronouns" in them.  Resumptives are pronouns that function in relative clauses much like gaps do in English.  They are incredibly common in the languages of the world; often they are just the ordinary personal pronouns put to this special purpose.  This is the case in non-standard English usage, as in the resumptive variant of the last ear-piercing relative above:

a guy who if he pierced his ears he would have gravy coming out of them

I'll call this sort of example, where the resumptive is in alternation with a gap, a ResPrince gapless relative -- Res for resumptive, Prince for Ellen Prince, who's studied them (see her 1990 article "Syntax and discourse: a look at resumptive pronouns", in BLS 16.482-97).  Another example, a paraphrase of Michael Moore speaking in his movie Fahrenheit 9/11, as discussed on the OutIL mailing list in October 2004:

Iraq, a country that it has never attacked us, that it has never threatened us...

The gapped variant is fine (and standard):

Iraq, a country that ___ has never attacked us, that ___ has never threatened us...

Still another example, from a NYT op-ed piece by Peter Guralnick on 8/11/07:

Or, as Jake Hess, the incomparable head singer for the Statesmen Quartet and one of Elvis's lifelong influences, pointed out: "Elvis was one of those artists, when he sang a song, he just seemed to live every word of it..."

This one has the extra complication that the relative clause is a "zero relative", with no relativizer, and also a "subject relative", in which the relativized element within the relative clause functions as the subject there.  Subject zero relatives like "There was a farmer had a dog" are non-standard (though they do occur).  Eliminating that non-standard feature still leaves us with a non-standard ResPrince case:

one of those artists who/that, when he sang a song, he just seemed to live every word of it

The gapped version is possible, and standard:

one of those artists who/that, when he sang a song, ___ just seemed to live every word of it

Prince wrote in e-mail at the time that in her 1990 article she

argued that [these resumptives] occur in either nonrestrictives or else (this kind of) restrictives with an indefinite head -- the two kinds of relatives where the relevant entity is evoked by the head alone. (I.e. where the hearer can retrieve or create the discourse entity as soon as the head is uttered, without having to wait for the relative clause, as one has to with definite head restrictives.)

She noted that such resumptives are perfectly standard in Yiddish, though they're non-standard in English.

The ResPrince examples differ from another use of resumptives in English -- to serve in place of gaps in positions from which "extraction" is barred, as in this example from an interview on NPR's Morning Edition on 3/19/07 (the interviewee is talking about Wal-Mart):

They have a billion dollars of inventory that they don't know where it is.

Extraction from inside adverbial subordinate clauses is generally barred (in the technical terminology of syntax, adverbial subordinate clauses are islands for extraction):

*They have a billion dollars of inventory that they don't know where ___ is.

Resumptives are non-standard, but in such cases they're much better than their gapped counterparts, which people usually find incomprehensible, or at least very hard to comprehend.  I'll call this sort of example a ResIsland gapless relative.

A problem in analyzing the data: there are some examples that have a pronoun in the relative clause which is anaphoric to the head of the clause but where there can be some doubt that this pronoun is actually resumptive; these might really be (still more) NoPro cases rather than ResIsland cases.  Here's Holden Caulfield in The Catcher in the Rye, quoted in the New Yorker of 3/14/05, p. 132:

What really knocks me out is a book that, when you're all done reading it, you wish the author that wrote it was a terrific friend of yours and you could call him up on the phone whenever you felt like it.

There's a fair amount of irrelevant detail here.  In particular, the second conjunct in the coordinate object of wish is beside the point.  Also, the it in the subordinate when-clause is not the issue here.  Eliminating this stuff leaves:

a book that you wish the author that wrote it was a terrific friend of yours

The question is now what the status of the remaining it (referring to the book) is.  Is it a resumptive pronoun, or just an ordinary anaphoric pronoun?  Certainly the pronoun is not omissible:

*a book that you wish the author that wrote ___ was a terrific friend of yours

But is that because something has been extracted from inside a relative clause (another kind of island), or because pronoun arguments are not normally omissible in English, as in the example below?

Here's a recent book.  *I wrote.  [meaning 'I wrote it']

Maybe the it is incidental to the matter, as would be suggested by the fact that the relative clause (with head the author) that it's in can be removed, leaving us with a pretty clear NoPro case:

a book that you wish the author was a terrific friend of yours

So maybe the original Caulfield example is a NoPro case too.

I have a few more of these.

A final note: once again, non-standard usage in English reflects syntactic patterns that are standard in other languages.  NoPro relatives are like relative types in (among other languages) Japanese and Korean; ResPrince relatives are like a relative type in (at least) Yiddish; and ResIsland relatives involve resumptive pronouns of a very ordinary sort -- deployed in English to allow expression of meanings that can't be easily expressed by gapped relatives, which are subject to constraints on extraction, while pronouns are not.

Non-standard (and regional and social) varieties are languages, period, and can be expected to exhibit phenomena found in varieties (standard or non-standard) of other languages.  As a result of this fact, studying a range of non-standard (etc.) varieties of one language can provide a rich set of data -- not unlike those derived from typological studies of the ordinary sort and fieldwork on little-known languages -- about what languages can be like.  This is a commonplace among linguists, but it's not always appreciated by other people.

Posted by Arnold Zwicky at October 14, 2007 04:02 PM